evaluation of analytical similarity between trastuzumab biosimilar … · 2019-09-06 · useful...
TRANSCRIPT
Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=kmab20
mAbs
ISSN: 1942-0862 (Print) 1942-0870 (Online) Journal homepage: http://www.tandfonline.com/loi/kmab20
Evaluation of analytical similarity betweentrastuzumab biosimilar CT-P6 and referenceproduct using statistical analyses
Jihun Lee, Hyun Ah Kang, Jin Soo Bae, Kyu Dae Kim, Kyoung Hoon Lee, KiJung Lim, Min Joo Choo & Shin Jae Chang
To cite this article: Jihun Lee, Hyun Ah Kang, Jin Soo Bae, Kyu Dae Kim, Kyoung Hoon Lee,Ki Jung Lim, Min Joo Choo & Shin Jae Chang (2018) Evaluation of analytical similarity betweentrastuzumab biosimilar CT-P6 and reference product using statistical analyses, mAbs, 10:4,547-571, DOI: 10.1080/19420862.2018.1440170
To link to this article: https://doi.org/10.1080/19420862.2018.1440170
© 2018 Celltrion Inc. Published with licenseby Taylor & Francis Group, LLC© CelltrionInc.
Accepted author version posted online: 26Feb 2018.Published online: 14 Mar 2018.
Submit your article to this journal
Article views: 1897
View Crossmark data
REPORT
Evaluation of analytical similarity between trastuzumab biosimilar CT-P6and reference product using statistical analyses
Jihun Lee , Hyun Ah Kang, Jin Soo Bae , Kyu Dae Kim, Kyoung Hoon Lee, Ki Jung Lim, Min Joo Choo,and Shin Jae Chang
Biotechnology Research Institute, R&D Division, Celltrion Inc., Incheon, Korea
ARTICLE HISTORYReceived 18 January 2018Revised 7 February 2018Accepted 9 February 2018
ABSTRACTThe evaluation of analytical similarity has been a challenging issue for the biosimilar industry because thenumber of lots for reference and biosimilar products available at the time of development are limited,whilst measurable quality attributes of target molecule are numerous, which can lead to potential bias orfalse negative/positive conclusions regarding biosimilarity. Therefore, appropriate statistical analyses arehighly desirable to achieve a high level of confidence in the similarity evaluation. A recent guideline forthe risk-based statistical approaches recommended by the US Food and Drug Administration providesuseful tools to systematically evaluate analytical similarity of biosimilar products compared with referenceproducts. Here, we evaluated analytical similarity of CT-P6, a biosimilar product of trastuzumab, with thereference products (EU-Herceptin� or US-Herceptin�) following these statistical approaches. Variousquality attributes of trastuzumab were first ranked based on the clinical impact of each attribute andsubsequently adjusted to one of three tiers (Tier 1, Tier 2 and Tier 3) considering the characteristics of theassay, the level of attribute present and the feasibility of statistical analysis. Two biological activities withhighest potential clinical impact were evaluated by an equivalent test (Tier 1), and other bioactivities andstructural/physicochemical properties relevant to the clinical impact were evaluated by a quality rangeapproach (Tier 2). The attributes with low risk ranking or qualitative assay were evaluated by visualcomparison (Tier 3). Analytical similarity assessment analyzed by the three tiers clearly demonstrated thatCT-P6 exhibits highly similar structural and physicochemical properties, as well as functional activities,compared with the reference products. There were small differences observed in a few quality attributesbetween CT-P6 and the reference products, but the differences were very minor, and unlikely to impact onclinical outcome. The recently reported equivalent clinical efficacy of CT-P6 with the reference productfurther supports that CT-P6 is highly similar compared with the reference product in the view of totality-of-evidence.
KEYWORDSbiosimilar; CT-P6;HerzumaTM; referenceproduct; statistical analyses;trastuzumab
Introduction
A biosimilar, also referred to as a follow-on-biologic, is a prod-uct that shows no clinically meaningful difference fromapproved biologics. Compared to developing generic of chemi-cal drug, developing biosimilar product is challenging due tolarge size and inherent complexity of biologics which is pro-duced by living cells. As the quality of biologics is process-dependent and most critical information for manufacturingand analysis remains confidential, developers must useadvanced technology from initial development to the final qual-ity assessment of the product to be marketed. Abbreviatedlicensure pathway for biosimilar is provided based on thenotion that biosimilarity has been demonstrated.1–4 Althoughdetails vary among countries, guidelines commonly indicatethat no difference in potential clinical impact in terms of safety,purity and potency should be shown between a biosimilarproduct and a reference product, using orthogonal and state-of-the-art assays that can fully identify the physico-chemicaland biological characteristics of the product. Accordingly,
several biosimilar products have shown their analytical biosimi-larity with reference products.5–10
CT-P6 was developed as a biosimilar of Herceptin�
(trastuzumab) by Celltrion Inc., and recently received a pos-itive opinion for market authorization by the EuropeanMedicines Agency (EMA)’s Committee for Medicinal Prod-ucts for Human Use.11 Herceptin� is approved for thetreatment of patients with certain breast and metastatic gas-tric cancers that overexpress human epidermal growth fac-tor receptor 2 (HER2) by the US Food and DrugAdministration (FDA) and EMA.12,13 Herceptin� has beenshown to inhibit the uncontrolled growth of cancer cells viabinding to the extracellular domain of HER2 in breast can-cers that express elevated levels of HER2.14 Also, like othermonoclonal antibodies (mAbs) used for cancer treatment, itcan kill cancer cells through antibody-dependent cell-medi-ated cytotoxicity (ADCC).15–18 Two other trastuzumab bio-similar products (Ontruzant�, developed by SamsungBioepis and OgivriTM, developed by Mylan and Biocon)
CONTACT Shin Jae Chang [email protected] Biotechnology Research Institute, R&D Division, Celltrion Inc. 23, Academy-ro, Yeonsu-gu, Incheon,South Korea 22014.© 2018 Celltrion Inc., Published with license by Taylor & Francis Group, LLC
MABS2018, VOL. 10, NO. 4, 547–571https://doi.org/10.1080/19420862.2018.1440170
have received approval recently in the European Union(EU) and United States (US), respectively.19,20
Because the biosimilarity is assessed in terms of totality-of-evi-dence, statistical evaluation for the assay is critical. It is well knownthat the original products exhibit lot-to-lot heterogeneity in certainquality attributes.21–24 Therefore, assessment should be performedwith an adequate number of meaningful lots and attributes/assaysshould be considered. For determining a statistical evaluationmethod, FDA recommended a risk-based approach.25 They sug-gested that the analytical similarity assessment should be plannedwith risk ranking of the reference product’s quality attributes basedon the potential impact on the clinical outcome, and then the statis-tical method should be applied among three tiers with appropriatesimilarity acceptance criteria. Equivalence testing (Tier 1) is appliedfor attributes with the highest potential clinical impact and assaysthat explain clinically relevant mechanism of action of the drug.Two one-sided tests on the mean difference of developing and thereference products are used, and use of an interval (¡1.5£sR,1.5£sR) that can support 90% confidence interval is recom-mended. A quality ranges approach (Tier 2) is recommended forattributes with lower risk rank. Standard deviation of the referenceproduct is used to determine similarity acceptance criteria, andthe multiplier for standard deviation should be justified based onthe scientific background. Evaluation through visual comparison(Tier 3) is recommended for attributes with the lowest risk rank-ing and test of qualitative results. According to FDA’s guideline,in addition to risk ranking, other factors such as level of attributes,tier of other orthogonal assays for the same attributes, and charac-ter of assays should be considered in determining tier.25
Herein, we evaluated the analytical similarity of CT-P6 tothe reference product Herceptin� in two regions (US and EU)following the tier approach. The results demonstrate that CT-P6 has highly similar structural/physicochemical propertiesand biological activities compared with EU-Herceptin� andUS-Herceptin�. Our experience shows that statistical evalua-tion using the tier approach is a useful tool to systematicallyevaluate analytical similarity results.
Results
As outlined in the various regulatory guidelines on the develop-ment of biosimilar products,1–4,26 a step-wise approach hasbeen taken with respect to the demonstration of similarity ofCT-P6 to Herceptin�, starting with a comprehensive physico-chemical and biological characterization of CT-P6 relative toits reference product. A wide range of state-of-the-art orthogo-nal methodologies were used to compare physicochemicalproperties and biological activities of CT-P6, EU- and US-Herceptin�. The investigated attributes include the primarystructure, higher order structure, purity/impurity profiles,charged variants, glycan structures, as well as various aspects ofproduct functionalities. The evaluation of the analytical similar-ity was conducted as described below.
Risk ranking
In order to score risk ranking based on the clinical impact of eachattribute, the guideline suggests consideration of at least two fac-tors: 1) potential impact of an attribute on clinical performance,
and 2) the degree of uncertainty around a certain quality attri-bute.25 Accordingly, we developed a risk assessment strategy withfive levels (very high ! high! medium ! low ! very low) of“potential clinical impact” and three levels (high ! medium !low) of “degree of uncertainty” (Table 1). Prior knowledge from lit-erature reports and the results of in-house studies were utilized todetermine the levels for these two factors. For example, the level of“potential clinical impact” selected was “very high” if an attributedirectly affects clinical outcome (e.g., potency, pharmacokinetics(PK)/pharmacodynamics (PD), safety, and immunogenicity). Thelevel of “degree of uncertainty”was determined to be “high” if thereis limited understanding of the clinical impact of an attribute. Riskrankings for each quality attribute were subsequently determinedby multiplying the scores of “potential clinical impact” and “degreeof uncertainty”. In general, biological activities were given greaterweight on the risk ranking than physicochemical properties sincethey directly measure activities linked to mechanisms of action,activity, efficacy, safety and immunogenicity of the product.
For example, risk ranking for anti-proliferation activity,HER2 binding affinity and ADCC activity were determined tobe “very high” since they are key mechanisms of action of tras-tuzumab for its clinical efficacy as a treatment for breast cancer.Other clinically relevant biological activities were ranked either“high” or “moderate” depending on their significance on clini-cal efficacy. Other biological activities such as C1q bindingaffinity were ranked “low” since they have little role in the anti-tumor activity of trastuzumab.
With regards to the physicochemical properties, primaryamino acid sequence that is directly related to efficacy andsafety of the product were ranked “high”. Purity related attrib-utes such as aggregates and fragments were ranked “moderate”considering they may cause reduced biological activity andimmunogenicity. Selected quality attributes, including heavychain (HC) Asp102 isomerization, non-glycosylation, afucosy-lation and high mannose, that have been known to affect bio-logical activity of the product were also ranked “moderate”.Risk ranking of the rest of the quality attributes such as C-ter-minal lysine variants were ranked “low” or “very low” sincethey have little impact on clinical efficacy or safety.
Tier classification
Based on the risk ranking, the level of tier was subsequentlyadjusted considering the characteristics of the assay, includingits sensitivity, the level of attribute present and the feasibility ofstatistical analysis.25 Tier 1 statistical analysis was carried outusing the equivalence test for quality attributes with “veryhigh” risk ranking that directly evaluate clinically relevantmechanisms of action (e.g., anti-proliferation or ADCC) of theproduct. Tier 2 statistical analysis was carried out using the
Table 1. Risk ranking determination and tier classification.
Potential Clinical Impact Degree of Uncertainty Risk Ranking Tier
Very High,High,
Very High 1
High,Medium,
High2
Medium, £Low
D ModerateLow, Low
3Very LowVery Low
548 J. LEE ET AL.
quality range approach to assess the quality attributes with the“high” to “moderate” risk ranking, for which quantitative datacan be obtained. Tier 3, an evaluation by raw or graphical datacomparison, was applied for the quality attributes with “low” to“very low” risk ranking. Tier 3 was also assigned for attributeswhere statistical analysis is not feasible. The strategy of tier clas-sification is shown in Table 1, and the assigned tier for eachquality attribute is listed in Table 2.
Similarity evaluation by Tier 1
Tier 1 equivalence testing is applied for attributes with thehighest potential clinical impact and assays that explain clini-cally relevant mechanism of action of the drug. In the qualityattributes of trastuzumab, anti-proliferation activity and ADCCactivity were selected as Tier 1 (Table 2).
Trastuzumab binds to the extracellular domain of HER2,and has been shown to selectively exert anti-tumor effects incancer models and patients with HER2-amplified breast cancer.Trastuzumab binding to HER2 blocks proteolytic cleavage ofthe extracellular domain of HER2, resulting in diminished lev-els of the more active p95–HER2 form of HER2 and selectivelyinhibiting ligand-independent HER2 interactions, includingHER2-HER3 heterodimerization, leading to cell cycle arrestand reducing proliferation.27–30 As a result of these effects onthe HER2 receptor, trastuzumab causes downregulation ofPI3K-Akt-mTOR pathway signaling and regulators of cell cycleprogression such as cyclin-dependent kinase inhibitor p27kip1and cyclin D1.30,31 As PI3K-Akt-mTOR signaling relate to pro-liferation, and tumor growth and invasion is a major effector ofHER2 activity, the blockade of the PI3K signaling pathway bytrastuzumab suppresses tumor growth in multiple models ofHER2-overexpressing breast cancer.32,33 Therefore, anti-prolif-eration activity plays a critical role in a mechanism of action oftrastuzumab in HER2-overexpressing cancer.
ADCC activity of trastuzumab is mediated by the interactionof the Fc region with Fcg receptors, which are expressed onvarious immune effector cells (macrophages/monocytes andnatural killer (NK) cells), but especially FcgRIIIa (CD16).34
ADCC associated with trastuzumab treatment has been dem-onstrated in numerous breast cancer cell lines.15,16 Thus,ADCC is considered to be a key mechanism of action of trastu-zumab in anti-tumor effects, and NK cells within peripheralblood mononuclear cells (PBMCs) play a key role in the ADCCeffect of trastuzumab.17,18–35
The results of the relative anti-proliferation activity and rela-tive ADCC activity of CT-P6, EU- and US-Herceptin� areshown in Table 3. Fig. 1A present scatter plots of the anti-pro-liferation activity and their equivalence test results comparingCT-P6 vs. EU-Herceptin� as well as CT-P6 vs. US-Herceptin�.The results demonstrate that the 90% confident interval (CI) ofmean difference between 2 products are within the equivalencemargin (EM), indicating that CT-P6 and Herceptin� are statis-tically equivalent and thus have highly similar potency in termsof anti-proliferation activity. Scatter plots of the ADCC activityfor CT-P6 and Herceptin� are shown in Fig. 1B. The equiva-lence test results revealed the 90% CI of mean differencebetween 2 products are within the EM, suggesting that CT-P6has statistically equivalent ADCC activity with Herceptin�.
Similarity evaluation by Tier 2
The Tier 2 quality ranges approach is used for the attributeswith lower risk rank. Summaries of analytical results and simi-larity evaluation for Tier 2 attributes for biological and physico-chemical assays are shown in Table 3 and Table 4, respectively.
With regard to the biological quality attributes, most of thebiological activity assays that do not directly assess mechanismsof action of trastuzumab were classified as Tier 2. A total of 8assays to analyze F(ab0) and Fc related functionalities were eval-uated by Tier 2.
To evaluate F(ab0) related activities, HER2 binding affinitywas measured by enzyme linked immunosorbent assay (ELISA)and cell-based ELISA (CELISA) methods. Although HER2binding was ranked “very high” because trastuzumab mediatesits activities through interaction with HER2,36–41 HER2 bindingassays by either ELISA or CELISA were classified as Tier 2 sincethey do not directly measure clinically relevant mechanisms ofaction of the product. Tier 2 analysis showed that all values ofHER2 binding affinity for CT-P6 were within the quality rangeof EU-Herceptin� and US-Herceptin� (Table 3, Fig. 2A andFig. 2B). These results suggest that CT-P6 exhibits highly simi-lar F(ab0) related activities compared with the referenceproducts.
For Fc-related activities, relative binding affinity toFcgRIIIa-V, FcgRIIIa-F, FcgRIIIb, FcgRIIa, FcgRIIb and FcRnwere measured by surface plasmon resonance (SPR). MAbsbound to a cell-surface antigen interact with FcgRs expressedon effector cells such as NK cells, neutrophils and macro-phages, inducing these cells to exert cytotoxicity.42 Among thehuman Fcg receptors (FcgRs), FcgRIIIa is well known as theonly activating FcgR expressed on NK cells, and can play a piv-otal role in ADCC induced by IgG1 subclass mAbs.34–43
FcgRIIIa is also found on the surface of resident monocytes,monocyte-derived dendritic cells and macrophages.44 FcgRIIareceptor has been reported to play a dominant role in the anti-body-dependent cellular phagocytosis activity of macro-phages.45,46 The binding to FcRn is considered an importantquality attribute related to the PK profile.47 Based on this priorknowledge from literature, similarity data of relative bindingaffinities to FcgRIIIa-V, FcgRIIIa-F, FcgRIIIb, FcgRIIa,FcgRIIb and FcRn were analyzed by a Tier 2 quality rangeapproach. Tier 2 analysis showed that all values of bindingaffinities to Fcg receptors and FcRn for CT-P6 were well withinthe quality range of EU-Herceptin� and US-Herceptin�
(Table 3, Fig. 2C to Fig. 2H). These results demonstrate thatCT-P6 exhibits similar Fc functionality compared with EU-and US-Herceptin�.
With regard to the physicochemical properties, selectedphysicochemical attributes that are quantifiable and have beenknown to potentially affect clinical efficacy and safety wereranked as Tier 2. The various assays analyzed for primary struc-ture, post-translational modification, higher order structure,purity/impurity, charge variants and glycosylation wereincluded in Tier 2 analysis (Table 2). The statistical analysisusing the quality range of the reference product revealed that,for most of Tier 2 attributes, more than 90% of data pointswere within the quality range of EU- and US-Herceptin�, asshown in Table 4.
MABS 549
Table 2. Quality Attributes and Test Methods used for Physicochemical and Biological Similarity Assessment between CT-P6 and Herceptin�.
Attribute Risk Ranking Test/Assay MeasurementTier of Statistical
Analysis
Primary Structure Primary Amino AcidSequence
High Molar Absorptivity Molar absorptivity/ Extinctioncoefficient
2
Peptide Mapping (HPLC) Peak profile 31
Peptide Mapping (LC-MS) Peak profile 31
N-terminal Sequencing Sequence identity 31
C-terminal Sequencing Sequence identity 31
Post Translational Deamidation Moderate Peptide Mapping (LC-MS) % Deamidation (Asn) 2Modifications Oxidation Low Peptide Mapping (LC-MS) % Oxidation (Met) 3
Isomerization Moderate Peptide Mapping (LC-MS) % Isomerization (Asp) 2N-terminal Glutamine Very Low Peptide Mapping (LC-MS) % N-terminal glutamine 3C-terminal Lysine
TruncationVery Low Peptide Mapping (LC-MS) % C-terminal lysine 3
C-terminal ProlineAmidation
Very Low Peptide Mapping (LC-MS) % C-terminal proline amidation 3
Higher OrderStructure
Disulfide Bonds Moderate Native and Reduced PeptideMapping
Disulfide bond positions 3
Free Thiol Analysis Free thiol (SH groups) 2Secondary Structure FTIR Structure of protein by spectroscopy 31
Secondary and TertiaryStructure
CD a-helical, b-sheet and unorderedstructures
31
Thermal Stability DSC Thermal unfolding temperatures 2Tertiary Structure Antibody Array 3D conformational epitope exposure 31
Purity/Impurity Aggregates Moderate SEC-HPLC % Monomer content 2% HMW Content 2
SEC-MALS Monomer size (kDa) 2HMW size (kDa) 32
% Monomer content 2% HMW content 2
AUC Monomer S value 2Dimer S value 2% Monomer content 2% Dimer content 2
Fragments Moderate Non-reduced CE-SDS % Intact IgG 2% Sum of non-Assembled Fragments 2
HCP Moderate Residual Host Cell Protein ELISA HCP (ppm) 3Host Cell DNA Moderate Residual Host Cell DNA PCR DNA (ppb) 3rProteinA Moderate Residual rProtein A ELISA rProtein A (ppm) 3Sub-visible Moderate MFI 1�, <100 (mm) 3particles 2�, <100 (mm)
5�, <100 (mm)10�, <100 (mm)15�, <100 (mm)25�, <100 (mm)40�, <100 (mm)50�, <100 (mm)70�, <100 (mm)(Cumulative Counts/mL)
Charge Variants Charge Variants Moderate (Deamidated/Isomerization)
IEF pI values of charge variants 31
Very Low (All others) IEC-HPLC % Peak 1CPeak 2CPeak 3CPeak 4 2% Peak 6 2% Peak 5 3% Peak 7
Glycosylation Non-glycosylatedProduct
Moderate Reduced CE-SDS % Non-glycosylated Heavy Chain %LCH
2
Afucosylated Glycans Moderate Oligosaccharide Profiling %G0CG1CG2 2N-linked Glycan Analysis %G0CG1CG2 2
High Mannose Glycans Moderate Oligosaccharide Profiling %Man5CMan6C Man8 2N-linked Glycan Analysis %Man5 2
Total AfucosylatedGlycans
Moderate Oligosaccharide Profiling %G0CG1CG2CMan5CMan6CMan8 2
N-linked Glycan Analysis %G0CG1CG2C Man5 2Galactosylated Low Oligosaccharide Profiling Agalactosylated: 3Glycans % G0F,
%G0F-GNGalactosylated: 3%[G1F-GN]CG1FCG2F
N-linked Glycan Analysis Agalactosylated: 3% G0FGalactosylation: 3%G1FCG2F
(Continued on next page)
550 J. LEE ET AL.
A number of techniques were employed to compare the pri-mary structure of CT-P6, EU- and US-Herceptin�, includingmolar absorptivity, peptide mapping (HPLC and LC-MS) andN/C-terminal sequencing. Although the risk ranking of pri-mary structure is “high”, most of methods are qualitative, andthus not amenable to quantitative evaluation. Molar absorptiv-ity and extinction coefficient provide quantitative data, so theywere analyzed by quality range analysis. It was found that themolar absorptivity and extinction coefficient values of CT-P6determined by measuring the optical density at a wavelength of280 nm and protein molarity derived from the amino acid anal-ysis were highly similar with EU- and US-Herceptin� (Table 4).The Tier 2 statistical analysis showed that 100% of the datapoints of molar absorptivity and extinction coefficient of CT-P6 were within the quality range of EU-Herceptin� as well asUS-Herceptin� (Table 4).
Peptide mapping by LC-MS was used to check primarystructure of the sample and to identify the post-translationalmodifications (deamidation, oxidation, isomerization, N-termi-nal pyroglutamate variants, C-terminal lysine variants and C-terminal proline amidation) of CT-P6, EU- and US-Herceptin�. Among the possible modifications, isomerizationat HC Asp102 (HC isoAsp102) and deamidation at light chain(LC) Asn30 were considered to potentially impact productpotency because they are located in the complementarity-deter-mining region-3.48 In particular, isomerization at HC Asp102has been known to severely reduce in-vitro anti-proliferationactivity of trastuzumab.49 Tier 2 statistical analysis showed thatlevels of HC isoAsp102 and deamidated LC Asn30 in CT-P6were within the quality range of EU-Herceptin� (Table 4),whereas, only 55.6% of data points for deamidated LC Asn30in CT-P6 were within the quality range of US-Herceptin�
(Table 4). This is due to the slightly lower deamidation level at
LC Asn30 in CT-P6 compared to US-Herceptin�; therefore, noadverse impact is expected due to the difference. Box plots ofthese modifications for the comparison between CT-P6, EU-and US-Herceptin� are shown in Fig. 3A and Fig. 3B.
In the higher order structure characterization, free thiol lev-els determined by Ellman assay and thermal transition temper-atures measured by differential scanning calorimetry (DSC)were evaluated using quality range analysis. HCs and LCs ofmAbs are linked by disulfide bonds, which contribute to struc-tural stability, thus free thiol content can reflect structuralintegrity of the product. The quality range evaluation suggestedthat 100% and 72.2% of CT-P6 lots were within the qualityrange (QR) of EU- and US-Herceptin�, respectively (Table 4).The 72.2% QR value of CT-P6 against US-Herceptin� origi-nated from lower free thiol content of CT-P6, indicatingslightly better structural integrity of CT-P6. DSC measures theheat capacity required to induce a change in the structure of amolecule, thus comparison of thermal transition temperaturesis useful in comparing the higher order structure of the prod-ucts. The Tier 2 statistical analysis showed more than 90% QRvalues in the three thermal transition temperatures correspond-ing to CH2, Fab, and CH3 unfolding. These results indicate thatthermal stability and conformation of CT-P6 are highly similarto EU- and US-Herceptin� (Table 4).
A wide range of methods were used to characterize thepurity and impurity levels of the products (Table 2). Aggrega-tion is a significant concern for biopharmaceutical productsbecause it may be associated with decreased bioactivity andincreased immunogenicity.50–54 Thus, the monomer and aggre-gate contents in CT-P6 and the reference products were thor-oughly examined by three orthogonal methods, size-exclusionchromatography-high performance liquid chromatography(SEC-HPLC), SEC-multi-angle light scattering (SEC-MALS)
Table 2. (Continued ).
Attribute Risk Ranking Test/Assay MeasurementTier of Statistical
Analysis
Sialic Acids Low Oligosaccharide Profiling %[G1F-GNCNANA]C[G1FCNANA]C[G2FCNANA]C[G2FC2NANA]
3
Sialic Acid Analysis NANA (sialic acid / protein, mol / mol) 3Glycation Low Glycation analysis % Glycation at light chain 3
% Glycation at heavy chainFab Binding HER2 Binding Very High HER2 Binding Affinity (ELISA) Relative HER2 Binding (%) 23
Cell-based HER2 Binding Affinity(CELISA)
23
Anti-proliferation Very high In Vitro Bioactivity (anti-proliferation) using BT-474 Cell
Relative Anti-proliferation (%) 1
Fc Binding C1q Binding Low C1q Binding (ELISA) Relative C1q Binding (%) 3FcgRIIIa Binding High FcgRIIIa V Type Binding Affinity
(SPR)Relative FcgRIIIa V Type Binding
Affinity (%)2
FcgRIIIa F Type Binding Affinity(SPR)
Relative FcgRIIIa F Type BindingAffinity (%)
2
FcgRIIIb Binding Moderate FcgRIIIb Binding Affinity (SPR) Relative FcgRIIIb Binding Affinity (%) 2FcgRIIa Binding High FcgRIIa Binding Affinity (SPR) Relative FcgRIIa Binding Affinity (%) 2FcgRIIb Binding Moderate FcgRIIb Binding Affinity (SPR) Relative FcgRIIb Binding Affinity (%) 2FcgRI Binding Low FcgRI Binding Affinity (SPR) Relative FcgRI Binding Affinity (%) 3FcRn Binding Moderate FcRn Binding Affinity (SPR) Relative FcRn Binding Affinity (%) 2
Fab –Fc MediatedActivities
ADCC Very High ADCC (PBMC) Relative ADCC Potency (%) 1
1Tier 3 was assigned because nature of the assays is qualitative despite of “high” or “moderate” risk ranking.2Tier 3 was assigned due to the trace amount of HMW content to precisely evaluate the molecular weight by MALS.3Tier 2 was assigned considering HER2 binding affinity does not measure the MoAs (anti-proliferation or ADCC activities) directly relevant to the clinical efficacy oftrastuzumab.
MABS 551
and analytical ultracentrifugation (AUC). The QR evaluationand box plots for three methods are shown in Table 4 andFig. 4, respectively. Tier 2 statistical analysis revealed that levelsof monomer and high molecular weight (HMW) species inCT-P6 were highly similar (100% QR) with those in EU- andUS-Herceptin� when analyzed by SEC-HPLC and SEC-MALS,whilst monomer and dimer levels obtained from AUC showed88.9% QR. The low QR values in monomer and dimer levels inAUC were due to the slightly lower dimer content of CT-P6than the reference products, indicating subtle improvement inaggregate content in CT-P6. Considering totality of the datafrom the three orthogonal methods, it can be concluded thatthere is no significant difference with respect to monomer andaggregate contents in CT-P6, EU- and US-Herceptin�.
Fragmentation of mAbs is another source of product-relatedimpurities that can potentially affect clinical efficacy, safety andPK profile because the presence of fragments may change theefficacy, reduced half-life, different biodistribution and safetyprofile of the product.55 Non-reduced CE-SDS was utilized todetermine the level of fragmented species. As shown in Table 4
and Fig. 5A, the level of fragments in CT-P6 (2.3 § 0.40%) wasfound to be moderately higher than that in EU-Herceptin�
(1.7 § 0.25%) and US-Herceptin� (1.9 § 0.25%). The qualityrange analysis also showed lower than 90% QR values in thefragmentation level. The higher fragment content in CT-P6leads to somewhat lower intact IgG level compared to the refer-ence products (Fig. 5B). To evaluate the impact of fragmenta-tion level on biological activities, anti-proliferation, ADCC,binding affinities to FcRn and FcgRIIIa-V were measured forCT-P6 samples containing various levels of fragmentation(Fig. 5E). The study results showed that up to approximately5% of fragmentation level did not cause significant impact onthe biological activities. Therefore, slightly higher fragmenta-tion level in CT-P6 is not anticipated to impact on clinical effi-cacy of the product.
The charge variants of mAbs are generated by commonmodifications such as oxidation, deamidation, isomerization,amination, cyclization, glycation, and the presence of C-termi-nal lysines.56 Depending on the locations and types of modifi-cation, charged isoforms may adversely impact on biological
Table 3. Summary of Similarity Assessment Results for Biological Assays.
Similarity Evaluation (EM1 or QR2)
Quality Attribute / Test Method Product Min – Max Mean § SD EU vs. CT-P6 US vs. CT-P6
Tier 1F(ab’) related Activities3 Anti-proliferation EU-Herceptin� 97 – 113 105 § 4.3 Within EM Within EM
CT-P6 94 – 105 101 § 2.8US-Herceptin� 98 – 113 105 § 5.7
Fab-Fc Mediated Activities3 ADCC (%) EU-Herceptin� 83 – 117 96 § 10.6 Within EM Within EMCT-P6 91 – 107 99 § 5.8US-Herceptin� 79 – 110 91 § 11.6
Tier 2F(ab’) Binding3 HER2 binding affinity (ELISA) EU-Herceptin� 92 – 103 97 § 3.3 100.0% QR 100.0% QR
CT-P6 94 – 106 100 § 4.0US-Herceptin� 95 – 112 100 § 4.9
Cell-based HER2 binding affinity(CELISA)
EU-Herceptin� 87 – 113 99 § 7.3 100.0% QR 100.0% QR
CT-P6 90 – 116 100 § 8.8US-Herceptin� 87 – 114 102 § 9.3
Fc Binding3 FcgRIIIa-V binding affinity (SPR) EU-Herceptin� 73 – 103 90 § 9.2 100.0% QR 100.0% QRCT-P6 90 – 102 96 § 3.4US-Herceptin� 71 – 110 87 § 11.1
FcgRIIIa-F binding affinity (SPR) EU-Herceptin� 70 – 105 91 § 11.6 100.0% QR 100.0% QRCT-P6 96 – 104 99 § 2.2US-Herceptin� 80 – 101 90 § 7.3
FcgRIIIb binding affinity (SPR) EU-Herceptin� 67 – 108 89 § 12.0 100.0% QR 100.0% QRCT-P6 87 – 105 97 § 4.4US-Herceptin� 66 – 106 83 § 13.7
FcgRIIa binding affinity (SPR) EU-Herceptin� 94 – 102 99 § 2.1 100.0% QR 100.0% QRCT-P6 98 – 103 100 § 1.4US-Herceptin� 94 – 100 97 § 2.2
FcgRIIb binding affinity (SPR) EU-Herceptin� 90 – 104 96 § 3.9 100.0% QR 100.0% QRCT-P6 95 – 102 99 § 2.0US-Herceptin� 91 – 103 95 § 3.9
FcRn binding affinity (SPR) EU-Herceptin� 93 – 105 99 § 3.1 100.0% QR 100.0% QRCT-P6 95 – 102 100 § 2.0US-Herceptin� 93 – 105 99 § 4.0
Tier 3Fc Binding3 C1q binding affinity (ELISA) EU-Herceptin� 89 – 113 100 § 6.2 Highly similar Highly similar
CT-P6 92 – 115 104 § 7.1US-Herceptin� 93 – 110 102 § 6.8
FcgRI binding affinity (SPR) EU-Herceptin� 95 – 102 99 § 2.3 Highly similar Highly similarCT-P6 93 – 102 98 § 2.4US-Herceptin� 95 – 101 97 § 1.9
1EM was determined as 1.5sR of Herceptin� data and results were determined as 90% CI of mean difference between two products.2The QR limits were set based on the range of the values obtained for the reference product variation, expressed as 3 times Standard Deviation (SD). High similarity wasconsidered to have been demonstrated if 90% of data points were within the QR of Herceptin� lots.3Relative potency (%) or binding affinity (%) in comparison to CT-P6 in-house reference standard.
552 J. LEE ET AL.
activities thus it is important to identify these charge variants inthe product. The charge variant profile of CT-P6 consists ofmultiple molecular variants such as charged glycans, lysine var-iants, deamidated forms and isomerized forms. The chargedisoforms are resolved to 7 peaks using ion exchange chroma-tography (IEC)-HPLC as shown in Fig. 6A. CT-P6, EU- andUS-Herceptin� exhibited similar IEC-HPLC peak distribution.A peak fractionation study identified that Peak 6 contains HCisoAsp102 modification, therefore exhibits substantiallyreduced anti-proliferation activity. The study also showed thatacidic peaks (Peak1, 2, 3 and 4) contain relatively higher levelsof deamidated LC Asn30. Therefore, Peak 6 and sum of acidicpeaks in IEC-HPLC were evaluated by Tier 2 statistical analysis.As shown in Fig. 6B and Fig. 6C, all data points of CT-P6 werewithin the quality range of EU- and US-Herceptin� with regardto the level of Peak 6 and acidic peaks. These results suggestthat the content of charge variants in CT-P6 is highly similarcompared with that in the reference products.
Glycosylation of mAbs plays an important role for comple-ment-dependent cytotoxicity (CDC) and ADCC function bymodulating the binding to the Fcg receptor,57,58 and affects theantibody conformation and thermal stability.59 Therefore, it isimportant to carefully establish risk ranking related to the gly-cosylation. Due to the complexity of the structures and the vari-ety of effects on the product, the glycoforms were assessed infive different glycan groups: 1) non-glycosylated product, 2)afucosylated glycans, 3) mannosylated glycans, 4) galactosy-lated glycans, and 5) sialylated glycans. Among these glyco-forms, non-glycosylation, afucosylation and mannosylationwere classified as Tier 2.
Non-glycosylation may critically impact product potency andclinical efficacy considering the crucial roles of glycosylation on amAb’s function. We have determined the level of non-glycosyla-tion of trastuzumab by reduced CE-SDS. The similarity assessmentby reduced CE-SDS showed that the level of non-glycosylated HCin CT-P6 (0.8 § 0.20%) was slightly higher than that in EU-Herceptin� (0.6 § 0.10%) and US-Herceptin� (0.7 § 0.14%)(Table 4 and Fig. 5C). The Tier 2 statistical analysis also revealed77.8% QR and 88.9% QR when EU- and US-Herceptin� qualityranges were applied, respectively. The slightly higher non-glycosyl-ation in CT-P6 leads to the slightly lower level of HC and LC sumas shown in Fig. 5D. Nevertheless, a spiking study using fully agly-cosylated CT-P6 sample showed that there is no adverse impact onvarious biological activities up to about 3% of non-glycosylation(Fig. 5F). These results suggest that a slightly higher level of non-glycosylated HC in CT-P6 is highly unlikely to have any clinicalimpact.
To determine the relative content of afucosylated andmannosy-lated glycans, oligosaccharide profiling by 2-aminobenzamide(2-AB)-labeled hydrophilic interaction liquid chromatography-ultra-performance liquid chromatography (HILIC-UPLC) wasperformed. The oligosaccharide profile showed that the types andproportions of the glycans were reasonably conserved between CT-P6 and the reference products (Fig. 7A). With regard to afucosy-lated glycans (e.g., G0, G1 and G2), an extensive body of literaturehas shown that IgG1 lacking core fucose has increased bindingaffinity for FcgRIIIa, and subsequently exhibits enhanced ADCCactivity.60–64 A in-house spiking study using highly afucosylatedCT-P6 also demonstrated strong positive correlation of afucosyla-tion level with FcgRIIIa binding affinity, as well as ADCC activity
Figure 1. Scatter plots and equivalence test results for Tier 1 quality attributes. Comparisons of “CT-P6 vs. EU-Herceptin�” and “CT-P6 vs. US-Herceptin�” are shown in leftand right, respectively. (A) Scatter plots (up) of relative anti-proliferation activity for CT-P6 (blue dot), EU-Herceptin� (orange dot) and US-Herceptin� (grey dot). Equiva-lence test results (bottom) are shown with 90% CI of mean difference between the 2 products. (B) Scatter plots (up) of relative ADCC activity for CT-P6 (blue dot), EU-Herceptin� (orange dot) and US-Herceptin� (grey dot). Equivalence test results (bottom) are shown with 90% CI of mean difference between the 2 products.
MABS 553
Table4.
SummaryofSimilarityAssessmentR
esultsforP
hysicochem
icalTests.
SimilarityEvaluatio
n1
QualityAttribute/T
estM
ethod
Product
Min–Max
Mean§
SDEU
vs.CT-P6
USvs.CT-P6
Tier2
Primarystructure
MolarAb
sorptivity
MolarAb
sorptivity
(L∙mol¡1∙cm
¡1)
EU-Herceptin
�203,104–227,138
213,172§
6,750.8
100.0%
QR
100.0%
QR
CT-P6
202,824–222,010
212,741§
5,024.8
US-Herceptin
�202,999–220,972
212,833§
4,626.5
Extin
ctionCoefficient
(L∙g
¡1∙cm
¡1)
EU-Herceptin
�1.40
–1.56
1.47
§0.047
100.0%
QR
100.0%
QR
CT-P6
1.40
–1.53
1.47
§0.035
US-Herceptin
�1.40
–1.52
1.47
§0.031
Post-translatio
nal
modificatio
nPeptideMapping
(LC-MS)
DeamidationatLC
Asn30(%
)EU
-Herceptin
�6.7–10.5
8.8§
0.94
100.0%
QR
55.6%QR
CT-P6
8.6–9.5
9.0§
0.26
US-Herceptin
�9.2–10.2
9.8§
0.31
Isom
erizationatHCAsp102
(%)
EU-Herceptin
�2.4–7.7
4.0§
1.25
100.0%
QR
100.0%
QR
CT-P6
2.7–6.5
4.2§
1.11
US-Herceptin
�2.7–7.9
4.8§
1.34
Highero
rder
structure
Free
ThiolA
nalysis
Average(free
SH/Ig
G,m
M/m
M)
EU-Herceptin
�0.25
–0.36
0.31
§0.029
100.0%
QR
72.2%QR
CT-P6
0.25
–0.35
0.28
§0.023
US-Herceptin
�0.29
–0.34
0.31
§0.015
DSC
(�C)
Tm1
EU-Herceptin
�71.0–71.4
71.2§
0.10
94.4%QR
100.0%
QR
CT-P6
70.9–71.3
71.1§
0.09
US-Herceptin
�71.0–71.5
71.1§
0.14
Tm2
EU-Herceptin
�81.0–81.4
81.1§
0.14
100.0%
QR
94.4%QR
CT-P6
81.0–81.4
81.1§
0.10
US-Herceptin
�81.0–81.2
81.1§
0.08
Tm3
EU-Herceptin
�82.9–83.2
83.0§
0.11
100.0%
QR
94.4%QR
CT-P6
82.9–83.2
83.0§
0.08
US-Herceptin
�82.9–83.1
82.9§
0.06
Purity/Impu
rity
SEC-HPLC
Monom
er(%
)EU
-Herceptin
�99.5–99.7
99.6§
0.09
100.0%
QR
100.0%
QR
CT-P6
99.6–99.7
99.7§
0.05
US-Herceptin
�99.5–99.8
99.6§
0.09
HMW
(%)
EU-Herceptin
�0.3–0.5
0.4§
0.09
100.0%
QR
100.0%
QR
CT-P6
0.2–0.3
0.3§
0.04
US-Herceptin
�0.2–0.5
0.4§
0.08
SEC-MALS
Monom
er(%
)EU
-Herceptin
�99.3–99.6
99.5§
0.09
100.0%
QR
100.0%
QR
CT-P6
99.4–99.7
99.6§
0.08
US-Herceptin
�99.4–99.7
99.5§
0.10
HMW
(%)
EU-Herceptin
�0.3–0.6
0.4§
0.08
100.0%
QR
100.0%
QR
CT-P6
0.2–0.4
0.3§
0.07
US-Herceptin
�0.2–0.5
0.4§
0.09
Monom
er(M
W,kDa)
EU-Herceptin
�137–142
139§
1.2
100.0%
QR
100.0%
QR
CT-P6
138–141
139§
0.9
US-Herceptin
�137–141
139§
1.4
AUC
Monom
er(%
)EU
-Herceptin
�97.4–98.2
97.8§
0.24
100.0%
QR
88.9%QR
CT-P6
97.2–98.3
97.8§
0.25
US-Herceptin
�97.5–97.9
97.7§
0.14
Dimer(%
)EU
-Herceptin
�1.8–2.6
2.2§
0.24
100.0%
QR
88.9%QR
CT-P6
1.7–2.5
2.1§
0.21
US-Herceptin
�2.1–2.5
2.3§
0.14
Monom
er(s-value)
EU-Herceptin
�5.1–5.3
5.2§
0.07
100.0%
QR
100.0%
QR
CT-P6
5.1–5.3
5.2§
0.06
US-Herceptin
�5.1–5.2
5.2§
0.05
554 J. LEE ET AL.
Dimer(s-value)
EU-Herceptin
�7.3–7.6
7.5§
0.09
100.0%
QR
88.9%QR
CT-P6
7.4–7.7
7.5§
0.09
US-Herceptin
�7.4–7.6
7.5§
0.06
CE-SDS(non-reduced)IntactIgG(%
)EU
-Herceptin
�98.0–98.8
98.3§
0.25
66.7%QR
66.7%QR
CT-P6
96.8–98.3
97.7§
0.39
US-Herceptin
�97.7–98.5
98.2§
0.23
Sum
offragments(%
)EU
-Herceptin
�1.2–2.1
1.7§
0.25
66.7%QR
72.2%QR
CT-P6
1.7–3.3
2.3§
0.40
US-Herceptin
�1.5–2.3
1.9§
0.23
Charge
variants
IEC-HPLC
AcidicGroup
(Peak1C
2C3C
4)(%
)EU
-Herceptin
�20.2–31.4
26.5§
3.09
100.0%
QR
100.0%
QR
CT-P6
26.3–31.1
28.3§
1.81
US-Herceptin
�27.0–31.3
29.1§
1.73
Peak
6(%
)EU
-Herceptin
�4.3–6.4
5.6§
0.85
100.0%
QR
100.0%
QR
CT-P6
3.2–5.0
4.2§
0.65
US-Herceptin
�4.2–
6.3
5.2§
0.84
Glycosylatio
nCE-SDS(reduced)
HCL
(%)
EU-Herceptin
�99.1–99.6
99.4§
0.11
83.3%QR
88.9%QR
CT-P6
98.8–99.5
99.3§
0.20
US-Herceptin
�99.1–99.6
99.3§
0.14
Non-glycosylatedheavychain
(%)
EU-Herceptin
�0.5–0.9
0.6§
0.10
77.8%QR
88.9%QR
CT-P6
0.5–1.2
0.8§
0.20
US-Herceptin
�0.4–0.9
0.7§
0.14
Oligosaccharide
Profilingby
2-AB
labeledHILIC-UPLC
Afucosylation(%
)EU
-Herceptin
�3.1–7.9
5.1§
1.14
100.0%
QR
100.0%
QR
CT-P6
4.6–6.6
5.3§
0.49
US-Herceptin
�3.2–6.7
4.7§
1.09
HighMannose
(%)
EU-Herceptin
�1.3–3.1
2.2§
0.52
100.0%
QR
100.0%
QR
CT-P6
2.1–3.4
2.6§
0.34
US-Herceptin
�1.7–4.1
2.3§
0.68
TotalA
fucosylatio
n(%
)EU
-Herceptin
�4.6–11.1
7.3§
1.59
100.0%
QR
100.0%
QR
CT-P6
6.8–10.0
7.9§
0.77
US-Herceptin
�4.9–10.8
6.9§
1.57
N-linked
Glycan
Analysisby
LC-M
SAfucosylation(%
)EU
-Herceptin
�6.2–10.7
8.6§
1.34
100.0%
QR
100.0%
QR
CT-P6
7.2–11.0
9.7§
0.86
US-Herceptin
�5.1–9.9
7.6§
1.87
HighMannose
(%)
EU-Herceptin
�1.1–2.4
1.6§
0.38
100.0%
QR
100.0%
QR
CT-P6
1.2–1.6
1.4§
0.12
US-Herceptin
�1.2–2.3
1.6§
0.33
TotalA
fucosylatio
n(%
)EU
-Herceptin
�7.5–13.2
10.2§
1.65
100.0%
QR
100.0%
QR
CT-P6
8.8–12.6
11.2§
0.88
US-Herceptin
�6.3–11.6
9.2§
1.81
Tier3
Primarystructure
PeptideMapping
(HPLC)
Peak
profile
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
PeptideMapping
(LC-MS)
Peak
profile
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
N-terminalsequ
encing
Sequ
ence
identity
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
C-term
inalsequ
encing
Sequ
ence
identity
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Post-translatio
nal
modificatio
nPeptideMapping
(LC-MS)
%Oxidatio
natHCMet255
EU-Herceptin
�1.5–2.5
1.8§
0.24
Highlysimilar
Highlysimilar
CT-P6
1.8–2.5
2.1§
0.25
US-Herceptin
�1.7–2.3
2.0§
0.21
(Continuedon
nextpage)
MABS 555
Table4.
(Continued).
SimilarityEvaluatio
n1
QualityAttribute/T
estM
ethod
Product
Min–Max
Mean§
SDEU
vs.CT-P6
USvs.CT-P6
%N-terminalpyroglutam
ateat
HCGlu01
EU-Herceptin
�1.2–2.1
1.7§
0.33
Highlysimilar
Highlysimilar
CT-P6
1.4–1.7
1.5§
0.10
US-Herceptin
�1.8–2.2
1.9§
0.14
%C-term
inalclippedlysine
EU-Herceptin
�97.9–99.0
98.6§
0.27
Highlysimilar
Highlysimilar
CT-P6
95.8–96.8
96.2§
0.26
US-Herceptin
�98.5–99.1
98.9§
0.20
%C-term
inalprolineam
idation
EU-Herceptin
�0.1–0.4
0.2§
0.12
C-term
inalprolineam
idation
inCT-P6ishigh
erthan
Herceptin
�
C-term
inalprolineam
idation
inCT-P6ishigh
erthan
Herceptin
�CT-P6
2.2–3.1
2.7§
0.23
US-Herceptin
�0.1–0.3
0.2§
0.04
Highero
rder
Disulfide
bond
positio
ning
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
structure
CT-P6
US-Herceptin
�
Second
arystructureby
FT-IR
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Second
aryandTertiarystructureby
CDEU
-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
3Depito
peexposureby
Antib
odyArray
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Purity/Impu
rity
SEC-MALS
HMW
size
(kDa)
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Residu
alHostC
ell
ProteinELISA
HCP
(ppm
)EU
-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Residu
alHostC
ellD
NA
PCR
HCP
(ppb
)EU
-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Residu
alrProteinA
ELISA
rProteinA(ppm
)EU
-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Sub-visibleparticlesby
MFI
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
Charge
variants
IEF
pIvalues
ofcharge
variants
EU-Herceptin
�Not
Applicable
Not
Applicable
Highlysimilar
Highlysimilar
CT-P6
US-Herceptin
�
IEC-HPLC
%Peak
5EU
-Herceptin
�63.2–72.1
66.4§
2.60
Highlysimilar
Highlysimilar
CT-P6
62.2–65.8
63.9§
1.13
US-Herceptin
�63.2–65.9
64.6§
1.00
%Peak
7EU
-Herceptin
�0.9–2.4
1.5§
0.39
Highlysimilar
Highlysimilar
CT-P6
2.9–4.0
3.6§
0.27
US-Herceptin
�0.9–1.5
1.2§
0.22
Glycosylatio
nGalactosylatedglycan
byHILIC-UPLC
%[G1F-GN]C
G1FCG
2FEU
-Herceptin
�19.3–47.8
35.3§
9.52
Highlysimilar
Highlysimilar
CT-P6
44.1–47.9
46.2§
1.15
US-Herceptin
�22.3–49.3
33.9§
10.45
556 J. LEE ET AL.
Galactosylatedglycan
byLC-M
S%G1FCG
2FEU
-Herceptin
�18.4–45.1
34.4§
8.21
Highlysimilar
Highlysimilar
CT-P6
42.1–47.1
44.7§
1.23
US-Herceptin
�20.1–45.5
33.0§
10.04
Sialicacidsby
HILIC-
UPLC
%[G1F-GNCN
ANA]C
[G1FCN
ANA]C
[G2FCN
ANA]C
[G2FC2
NAN
A]
EU-Herceptin
�0.3–2.2
1.2§
0.55
CT-P6contains
high
ersialic
acidscomparedto
Herceptin
�
CT-P6contains
high
ersialic
acidscomparedto
Herceptin
�CT-P6
2.1–3.3
2.9§
0.36
US-Herceptin
�0.5–1.7
1.1§
0.41
SialicacidsAn
alysis
NAN
A(sialicacid/p
rotein,
mol/m
ol)
EU-Herceptin
�0.03
–0.06
0.05
§0.010
CT-P6contains
high
erNAN
Acomparedto
Herceptin
�CT-P6contains
high
erNAN
Acomparedto
Herceptin
�CT-P6
0.12
–0.14
0.13
§0.004
US-Herceptin
�0.03
–0.06
0.04
§0.010
Glycatio
nby
intact
mass
%Glycatio
natlight
chain
EU-Herceptin
�2.0–2.5
2.3§
0.16
Highlysimilar
Highlysimilar
CT-P6
2.3–2.7
2.6§
0.10
US-Herceptin
�2.0–2.5
2.3§
0.15
%Glycatio
natheavychain
EU-Herceptin
�3.1–3.7
3.4§
0.20
Highlysimilar
Highlysimilar
CT-P6
4.0–4.7
4.4§
0.18
US-Herceptin
�2.8–3.8
3.4§
0.39
1 IntheTier2assays,the
QRlim
itsweresetb
ased
ontherang
eof
thevalues
obtained
forthe
referenceproductvariatio
n,expressedas
3tim
esStandard
Deviatio
n(SD).Highsimilaritywas
considered
tohave
been
demonstratedif
90%ofdatapointswerewith
intheQRof
Herceptin
�lots.
MABS 557
(data not shown). Mannosylated glycans do not contain fucose;therefore, it has been reported that high mannose glycoformsexhibit higher FcgRIIIa binding and ADCC activity.65–67 However,it has been shown that mannosylated glycans induce fast clearanceof IgG1 via amannose receptormediatedmechanism.68,69 The sim-ilarity assessment by oligosaccharide profiling showed that the levelof afucosylation and high mannose in CT-P6 were similar withones observed in EU- and US-Herceptin� (Table 4, Fig. 7B andFig. 7C). Total afucosylation level was also found to be highly simi-lar between the products (Fig. 7D). The Tier 2 statistical analysisrevealed all data points in CT-P6 were within the quality range ofEU- and US-Herceptin� for afucosylation, highmannose and total
afucosylation levels (Table 4). Consistent results were obtained byN-linked glycan analysis by LC-MS (Table 4). Taken together, itwas concluded that CT-P6 contains highly similar afucosylatedand mannosylated glycans compared with EU- and US-Herceptin�.
Similarity evaluation by Tier 3
Tier 3, evaluation through visual comparison, is used for attrib-utes with lowest risk ranking and test of qualitative results.Summary of analytical results and similarity evaluation for Tier3 attributes are shown in Table 3 and Table 4.
Figure 2. Comparison of Tier 2 biological attributes for CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey). Box plots of relative (A) HER2 binding affinity byELISA, (B) cell-based HER2 binding affinity by CELISA, (C) FcgRIIIa-V binding affinity by SPR, (D) FcgRIIIa-F binding affinity by SPR, (E) FcgRIIIb binding affinity by SPR , (F)FcgRIIa binding affinity by SPR, (G) FcgRIIb binding affinity by SPR and (H) FcRn binding affinity SPR. Orange and grey broken lines represent quality range of EU-Herceptin� and US-Herceptin�, respectively. Box plot shows the interquartile range (box), median (band inside of box), maximum and minimum values (whiskers).
558 J. LEE ET AL.
In the biological activity assays, C1q binding affinity byELISA and FcgRI binding affinity by SPR were ranked as Tier 3due to their low risk ranking scoring. Table 3 showed that C1qbinding affinity and FcgRI binding affinity of CT-P6 werehighly similar compared with EU- and US-Herceptin�.
In the physicochemical tests, a number of qualitativemethods were evaluated through visual comparison. For pri-mary structure characterization, peptide mapping by HPLCand LC-MS were conducted. UV-based tryptic peptide map-ping revealed a highly similar peak profile among the
Figure 3. Box plots of (A) HC isoAsp102 and (B) deamidated LC Asn30 levels for CT-P6 (blue), EU-Herceptin� (orange) and USHerceptin� (grey). Orange and grey brokenlines represent quality range of EU-Herceptin� and US-Herceptin� , respectively. Box plot shows the interquartile range (box), median (band inside of box), maximumand minimum values (whiskers).
Figure 4. Comparison of monomer and aggregate contents for CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey). Box plots of (A) monomer % measured bySEC-HPLC, (B) HMW % measured by SEC-HPLC, (C) monomer % measured by SEC-MALS, (D) HMW % measured by SEC-MALS, (E) monomer % measured by AUC and (F)dimer % measured by AUC. Orange and grey broken lines represent quality range of EU-Herceptin� and US-Herceptin�, respectively. Box plot shows the interquartilerange (box), median (band inside of box), maximum and minimum values (whiskers).
MABS 559
products without missing or additional new peaks, and withcomparative retention time of each peak (Fig. 8). Total ionchromatograms of the tryptic peptide map obtained by LC-MS were also highly similar between CT-P6 and the refer-ence products (data not shown). N- and C-terminalsequence determined by peptide mapping LC-MS werefound to be same between the products. These data supportthe conclusion that primary structure of CT-P6 is identicalwith EU- and US-Herceptin�.
Among the post-translational modifications analyzed bypeptide mapping LC-MS, oxidation at HC Met255, N-termi-nal pyroglutamate at HC Glu01, C-terminal clipped lysineand C-terminal proline amidation at HC were ranked Tier3, considering their relatively low risk ranking. The resultsshowed that all above modification levels were highly simi-lar between CT-P6 and the reference products except for C-terminal proline amidation (Table 4). C-terminal prolineamidation was found to be higher in CT-P6 compared tothat in the reference products, with a range of 2.2 – 3.1%for CT-P6 and 0.1 – 0.4% for the reference products. The
proline amidation at C-terminal is located away from func-tionally related (Fc or Fab) regions. Also, C-terminal pro-line amidation has been reported to have no impact on Fceffector function, Fab binding activity and the productsafety of an IgG1 antibody.70,71 Overall, the three productsshowed highly similar biological activities, suggesting thatthe observed difference in C-terminal proline amidation hasno impact on efficacy in vivo.
For higher order structure characterization, various qual-itative methods were utilized. Disulfide bond positions bymeans of native and reduced peptide mapping coupled withLC-MS revealed eight new peaks in the native peptide mapthat were identified as disulfide bond linked peptide basedon the MS and MS/MS sequence analysis. These eight disul-fide bond linked peptides were matched in all products,indicating the presence of the same inter- and intra-chaindisulfide bond linages in all products. The secondary or ter-tiary structures of the molecule were analyzed by circulardichroism (CD) and Fourier transform infrared spectros-copy (FT-IR). The far-UV and near-UV CD spectra of CT-
Figure 5. Comparison of fragmentation and non-glycosylation levels determined by CE-SDS for CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey). Box plotsof (A) sum of fragments % by non-reduced CE-SDS, (B) intact IgG % by non-reduced by CE-SDS, (C) non-glycosylated heavy chain % by reduced CE-SDS and (D) sum oflight and heavy chain % by reduced CE-SDS. Orange and grey broken lines represent quality range of EU-Herceptin� and US-Herceptin� , respectively. Box plot showsthe interquartile range (box), median (band inside of box), maximum and minimum values (whiskers). Effects of fragmentation and non-glycosylation level on biologicalactivities were examined in (E) and (F), respectively.
560 J. LEE ET AL.
P6, EU- and US-Herceptin� showed the typical shape of anantibody with highly similar spectral signal and mean resi-due molar ellipticity (MRE) (Fig. 9A and Fig. 9B). In FT-IRspectra, all samples agreed well with respect to shape andlocation of the amide I range of 1,641 to 1,642 cm¡1, theamide II band at the range of 1,553 to 1,554 cm¡1 and the3 characteristic bands between 1,200 and 1,500 cm¡1 (datanot shown). Antibody conformational array of the threeproducts showed highly similar epitope exposure in variableand constant regions against the 34 pools of polyclonalantibodies (Fig. 9C and Fig. 9D). These data indicated thatthe three products have highly similar secondary and ter-tiary structures.
The levels of process-related impurities, including residualhost cell protein, host cell DNA, residual rProtein A and sub-visible particles, were found to be highly similar between CT-P6 and the reference products. Charge variants analyzed by iso-electric focusing (IEF) and IEC-HPLC exhibited highly similarresults between the products.
With regard to the glycosylation, galactosylated glycans,sialic acid and glycations were compared between CT-P6and the reference products. Galactosylated glycans havebeen known to affect CDC activity,72,73 but their risk rank-ing was determined to be low because CDC is not a mecha-nism of action by which trastuzumab exerts its therapeuticeffect.74 The oligosaccharide profiling results showed CT-P6has a higher mean value of galactosylated glycan level
(46.2 § 1.15%) than EU-Herceptin� (35.3 § 9.52%) andUS-Herceptin� (33.9 § 10.45%) (Table 4). Nevertheless, itshould be noted that EU- and US-Herceptin� exhibit awide range of galactosylation levels; thus, all CT-P6 lotsremained within the minimum and maximum valuesobserved in EU- and US-Herceptin�. Chinese hamsterovary (CHO) cell-expressed IgG is known to contain onlytrace amount of N-acetylneuraminic acid (NANA), which isnot immunogenic.75 We consistently found that very lowlevels of NANA were found in CT-P6 and Herceptin� thatare expressed in the CHO cell line. Oligosaccharide profil-ing and sialic acid analysis showed that NANA levels inCT-P6 were higher than those of EU- and US-Herceptin�.However, the marginally higher amount of NANA presentin CT-P6 is not expected to cause adverse effects becausethe absolute amount of NANA present in CT-P6 remainsvery small and NANA is a non-immunogenic glycan spe-cies. Glycation is a post-translational chemical reaction thatresults in non-enzymatic addition of glucose to proteinamine groups, primarily the alpha amino terminal and epsi-lon amino group on the lysine side chain.76 The impactthat glycation has on therapeutic antibodies is somewhatcontroversial due to the fact that the effect of glycation isprotein specific, depending on exposed lysine residues. Gly-cation levels at LC and HC determined by intact massshowed that CT-P6 contains similarly low levels of glyca-tion compared with EU- and US-Herceptin� (Table 4).
Figure 6. Comparison of charge variants of CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey) analyzed by IEC-HPLC. (A) Representative ion exchange chro-matograms are presented for 3 batches of each product. The number and distribution of IEC-HPLC peaks are conserved between CT-P6 and RMPs. (B) Box plots of acidicpeaks % (Peak 1 C Peak 2 C Peak 3 C Peak 4) in IEC-HPLC, (C) Box plot of Peak 6 % in IEC-HPLC. Orange and grey broken lines represent quality range of EU-Herceptin�
and US-Herceptin�, respectively. Box plot shows the interquartile range (box), median (band inside of box), maximum and minimum values (whiskers).
MABS 561
Discussion
CT-P6 is being developed by Celltrion Inc. as a biosimilar prod-uct to the original trastuzumab, Herceptin�. A comprehensiveanalytical similarity assessment was performed using a widerange of sensitive, orthogonal and state-of-the-art methodologiesto demonstrate similarity of CT-P6 to the reference productHerceptin� obtained from the two different markets, EU andUS. The similarity assessment was performed to characterize pri-mary structure, higher order structure, purity/impurity, chargeisoforms, glycosylation and biological activities of trastuzumab.
The analytical data from the similarity study was evalu-ated using the statistical analyses recommended by FDA inits recent draft guideline. In this risk-based approach, vari-ous quality attributes of products are first ranked accordingto their risk of potential clinical impact. These attributes orassays are then evaluated by one of three tiers of statisticalanalysis based on a risk ranking, as well as other factors.Evaluation of analytical similarity has been a challengingissue for the biosimilar industry because the availability oflots for reference product and biosimilar product are lim-ited at the time of development, while the measurable
Figure 7. Comparison of oligosaccharide profiles of CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey) analyzed by HILIC-UPLC with 2-AB labeling. (A) Repre-sentative oligosaccharide profiles are presented for 3 batches of each product. The types and proportions of the glycans are conserved among the products. (B) Box plotof afucosylated glycans % (G0CG1CG2), (C) Box plot of high mannose glycans % (Man5CMan6CMan8), (D) Box plot of total afucosylated glycans %(G0CG1CG2CMan5CMan6CMan8). Orange and grey broken lines represent quality range of EU-Herceptin� and US-Herceptin�, respectively. Box plot shows the inter-quartile range (box), median (band inside of box), maximum and minimum values (whiskers).
562 J. LEE ET AL.
quality attributes of target molecule are numerous, whichcan lead to potential bias or false negative/positive conclu-sions about biosimilarity. Therefore, the risk-based tierapproach provides a systematic pathway to evaluate analyti-cal similarity with a high level of confidence and lowerpotential bias.
The data from our extensive similarity exercise success-fully demonstrated that CT-P6 is highly similar to EU-Herceptin� and US-Herceptin�. Similarity of anti-prolifera-tion and ADCC, the two biological activities with highestpotential clinical impact, were evaluated by the equivalenttest (Tier 1). The results demonstrated that 90% CI of
Figure 8. Comparison of UV chromatograms of trypsin digested CT-P6 (blue), EU-Herceptin� (orange) and US-Herceptin� (grey). Representative tryptic peptide mapsdetected at 214 nm are presented for 3 batches of each product.
Figure 9. Higher order structure analyzed by CD and antibody conformational array for CT-P6, EU-Herceptin� and US-Herceptin� . CD spectra for (A) Far-UV region and (B)Near-UV region are overlaid. Histograms of epitope exposure analyzed by antibody conformational array in (C) variable region, and (D) constant region are compared for 3batches of each product.
MABS 563
mean difference between CT-P6 and the reference productis within the equivalence margin in the anti-proliferationand ADCC assays. Most of the other biological activitiesrelated to F(ab0) or Fc functions were evaluated by the qual-ity range approach (Tier 2). The measured binding affinitiesof all CT-P6 lots were within the quality range of EU- andUS-Herceptin�. These results strongly suggest that CT-P6exhibits highly similar functional properties compared withthe reference products. Similarity of the structural andphysicochemical properties in trastuzumab were evaluatedby either Tier 2 or visual comparison (Tier 3) dependingon their risk ranking. Tier 2 and Tier 3 analyses showedhighly similar primary/higher order structures, purity/impu-rity profiles, charge isoforms and glycosylation between CT-P6 and the reference products. In the Tier 2 analysis, morethan 90% of CT-P6 lots compared were within the qualityrange of the reference product for most quality attributes.Although slightly higher levels of fragmentation and non-glycosylation were observed in CT-P6, the differences werevery small and unlikely to be clinically meaningful. Theattributes with low risk ranking or qualitative assay exam-ined by visual comparison also showed highly similarresults between CT-P6 and the reference product. Theseresults support that CT-P6 has highly similar structural/physicochemical properties compared with the referenceproducts. A recently reported clinical outcome study furthersupports comparable efficacy and safety of CT-P6 withHerceptin�.77,78 Taken together, it can be confidently con-cluded that CT-P6 is a highly similar product comparedwith the reference product in view of the totality of theevidence.
Materials and methods
Materials
CT-P6 were manufactured at Celltrion Inc., Korea. Herceptin�
lots were purchased from a pharmacies located in the EU andUS.
Tier 1 statistical analysis
Tier 1 statistical analysis uses the equivalence test with the nullhypothesis;
� H0: mT – mR � - d or mT – mR� d� H1: - d < mT – mR < d
Where mT stands for mean of tested product; mR stands formean of the reference product; and d stands for equivalencemargin (EM) based on the reference product variability (stan-dard deviation). The EM for demonstrating similarity wasdefined as § 1.5�Standard Deviation (SD) based on the refer-ence product. The value of 1.5 was established by the FDA.25
The confidence interval (CI) was used to determinewhether the mean difference for functional biological meas-ures from T (test or proposed biosimilar) and R (reference)products are similar or not. Prior to calculating the CI toassess similarity, homoscedasticity among products waschecked using an F-test, which is used to test if the varian-ces of products are equal. The p-value gives the probability
of a greater F value under the null hypothesis that popula-tion variances are equal against the alternative that the twoare different. F-test statistics is:
FD MAX s2T ; sR
2� �
MIN s2T ; sR
2� �
Where nT stands for number of test products; nR stands fornumber of the reference products; sT
2 stands for variance oftest products; sR
2 stands for variance of the reference productsthe null hypothesis is rejected if F > Fa 6 2;df 1;df 2 (whereFa 6 2;df 1;df 2 is upper tail critical value for the Fdf 1;df 2 distribu-tion). If sT
2 > sR2 then df 1D nT ¡ 1 and df 2D nR ¡ 1 : if sT
2 <
sR2 then df 1D nR ¡ 1 and df 2D nT ¡ 1:If the result of the F-test is not significant, implying equal
variance, the 90% confidence interval on the difference is calcu-lated using pooled standard deviation as below.
� (100(1-2a))% Confidence Interval of Mean Difference
.mT ¡ mR/§ t1¡a;df �
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffis2p
1nT
C 1nR
� �s
Where sp2 D .nT ¡ 1/s 2
T C .nR ¡ 1/s2R
nT C nR ¡ 2 and df � D nT C nR ¡ 2If the result of the F-test is significant, implying unequal var-
iance, 90% confidence interval on the difference is computedassuming the Satterthwaite method using the following formula
� (100(1-2a))% Confidence Interval of Mean Difference
mT ¡ mRð Þ§ t1¡a; df �
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffis2T
nTC s2
R
nR
sdf � D
s 2T
nTC s 2
RnR
� �2
s 4T
nT 2.nT ¡ 1/ C s 4R
nR2.nR ¡ 1/
Equivalence between 2 products was confirmed if 90% CI ofmean difference was within the corresponding EM. Equivalencetest was calculated by T-TEST PROCEDURE in SAS EnterpriseGuide 7.1 (SAS Institute Inc., Cary, NC, USA).
Tier 2 statistical analysis
Tier 2 statistical analysis uses the quality range of the referenceproduct:
� mR ¡ xsR; mR C xsRð ÞWhere mR stands mean of the reference product; sR stands
for variation (standard deviation) of the reference product; andx stands for multiplicity of unit reference product variation(multiplier).
The quality range (QR) was set based on the range of thevalues obtained from the reference product variation expressedas x times the reference product standard deviation (§xSD).The multiplier x has been determined for each Tier 2 attributebased on the variability of each assay and the relative impor-tance of the attribute to the safety and efficacy of the product.The multiplier 3 was used for all Tier 2 attributes. Based onFDA criteria, high similarity was considered to have been
564 J. LEE ET AL.
demonstrated if 90% of data points for the test product werewithin the QR of the reference products.79,80
Tier 3 evaluation
The raw data/graphical data are presented to allow comparisonof quality attributes with a low risk ranking or for qualitativetest methods.
Sample size calculation
The sample size for the similarity assessment was calculatedbased on assay variation. The variation of 6 batches each ofUS-Herceptin�, which were obtained from the initial simi-larity study, were evaluated with respect to biological activi-ties representative of in vitro bioactivity and ADCC as Tier1. The number of batches required to obtain an overall sta-tistical power of 90% were calculated using an EM defined as§ 1.5 � Standard Deviation (SD of US-Herceptin�) as this isconsidered as the most conservative statistical approach forassessment of similarity. A statistical sample-size calculationprogram, PASS ver.13.0 was used.81 Following this approach,the largest sample size was calculated to be 11 samples forthe assays. Therefore, more than 11 batches of each productwere included in the similarity assessment. A total 18 batchesof CT-P6, 17 batches of EU-Herceptin� and 12 batches ofUS-Herceptin� were used for the similarity assessment.
Amino acid analysis and molar absorptivity
After buffer exchange with phosphate-buffered saline, all sam-ples were hydrolyzed with 6 N HCl (including »1% phenol)under reduced pressure at 112�C for a 24 hrs. Decomposedamino acids were reconstituted and derivatized by pre-columnderivatization method with 6-aminoquinolyl-N-hydroxysucci-nimidyl carbamate (AQC reagent) so that the sample could bedetected by fluorescence detector. Each amino acid was quanti-tated with internal standard method.
Molar absorptivity was determined according to Beer-Lam-bert Law. First, measurements of OD280 and OD320 were per-formed using a Beckman DU730 spectrophotometer. Then,protein concentration was obtained from the amino acid analy-sis using the concentration of ‘robust’ amino acids (where %deviation between observed and expected results was � 5%)such as aspartic acid, glycine, arginine, alanine, proline, valineand leucine. Molar extinction coefficient was calculated withUV absorbance at 280 nm, concentration of protein and molec-ular weight of CT-P6 and Herceptin�.
HPLC peptide mapping analysis
Samples were analyzed by HPLC peptide mapping after reduc-tion with dithiothreitol (DTT) (Sigma-Aldrich), alkylation withiodoacetamide (IAM) (Sigma-Aldrich), and digestion withtrypsin (Promega) at 37�C. The resulting peptides were sepa-rated by reversed phase-HPLC using a C18 column (5 mm, 250£ 4.6 mm; Vydac, Hesperia) and acetonitrile gradient (Burdick& Jackson) containing trifluoroacetic acid (Sigma-Aldrich).Absorbance was monitored at 214 and 280 nm using a Waters-
2695 Alliance HPLC system equipped with a UV detector(Waters).
LC-MS-ESI peptide mapping analysis
Samples were analysed by LC-MS peptide mapping after reduc-tion with DTT, alkylation with IAM, desalting with NAP5 col-umn and digestion with trypsin (a 20 (sample):1 (trypsin)treatment, at 37�C for 2 hrs), Asp-N (a 100 (sample):1 (Asp-N)treatment, at 37�C for 16 hrs). The resulting peptides were sep-arated by reversed-phase UPLC using a C18 column and a gra-dient of acetonitrile containing formic acid. Thechromatography was performed on a Waters Acquity UPLC.An online AB SCIEX Triple TOF 5600 mass spectrometer withan electrospray source was used to collect mass spectra of theintact peptide, as well as to fragment the peptides for sequenc-ing (MS/MS analysis). The m/z (mass/charge) data were col-lected from 250 to 1,600 m/z.
Free thiol analysis by Ellman assay
The free thiol (SH) groups in the samples were determined bymeans of the 5,50-dithiobis-(2-nitrobenzoic acid) (DTNB, Ell-man’s reagent) method. Briefly, standard and samples weremixed with DTNB in 7.5 M guanidine HCl, 125 mM sodiumphosphate pH 8.0 and 1.25 mM EDTA, followed by measure-ment of absorbance at 412 nm. Free thiol groups could be esti-mated in a sample by comparison to a standard curvecomposed of known concentrations of a sulfhydryl-containingcompound such as cysteine. The results were reported as molarratios (free SH/IgG, mM/mM).
Disulfide bond analysis
Samples were analyzed by comparing native and reduced pep-tide maps. In the case of reduced peptide mapping analysis, thesamples were reduced with DTT and alkylated with IAM;whereas in the case of native peptide mapping analysis, noDTT was added to the sample. The samples were digested usingtrypsin after desalting with NAP-5 column. The resulting pepti-des were separated by reversed-phase UPLC using a C18 col-umn and a gradient of acetonitrile containing formic acid. Thechromatography was performed on a Waters Acquity UPLC.An online AB SCIEX Triple TOF 5600 mass spectrometer withan electrospray source was used to collect mass spectra of theintact peptide, as well as to fragment the peptides for sequenc-ing (MS/MS analysis).
Fourier transform infra-red spectroscopy
FT-IR spectra were recorded on a Nicolet 6700 (Thermo-Nico-let) with a smart iTR ATR accessory and diamond crystal at aresolution of 4 cm¡1, number of scans 32, and spectral rangefrom 4000 – 650 cm¡1. All spectra were baseline and ATR-cor-rected with instrument software. In order to generate the differ-ence spectra, buffer spectra were recorded as a blank andsubsequently subtracted. Sample solutions were transferredonto the crystal and allowed to dry. FT-IR spectra were ana-lyzed by comparing the locations and shapes of the amide I and
MABS 565
amide II bands, as well as three other bands between 1200 and1800 cm¡1.
Circular dichroism spectroscopy
CD experiments were performed using a Chirascan-plus CDspectrometer (Applied Photophysics) equipped with a Pelltiercontrolled temperature regulator at 20�C. Cells of quartz glassalong with optical path lengths of 1.0 and 0.10 cm were usedfor near-UV and far-UV CD measurements, respectively. Pro-tein concentrations used for the far-UV spectra and near-UVCD spectra were 0.2 and 1.0 mg/ml, respectively. The formula-tion buffer was measured as a blank and subsequently sub-tracted. Noise reduction was applied to the baseline-correctedprotein spectra using the smoothing option of the device soft-ware for the spectrometer. Conversion of the measured CD sig-nals to mean residue molar ellipticities [(u)MRE] was alsoperformed using this software.
Differential scanning calorimetry
Thermal stability of the samples was evaluated by measuringtheir Tm values using a Microcal VP-DSC microcalorimeter(MicroCal). The thermogram was obtained with a scan rate of1�C/min within a temperature range of 30 to 100�C. DSC datawere analyzed using a non-two state, 3 transition model to getthe melting points of the transition temperatures using the Ori-gin software package (OriginLab Corporation) to determinethree thermal transition temperatures.
Antibody array
CT-P6, EU-Herceptin� and US-Herceptin� samples weretested pairwise in triplicate using a HercBridge ELISA kit(ArrayBridge Inc., Cat. No. AB000207,). The assay was per-formed by making a 5 mg/ml solution of both CT-P6 andHerceptin� respectively, and adding to the 96-well plate.Following a 1 hour incubation to allow capture of the pro-teins by the panel of antibodies on the plate, a reportingpolyclonal anti-human IgG antibody, conjugated with bio-tin, was added and incubated for 1 hour to allow it to bindto any captured proteins. After this incubation, the platewas washed and a streptavidin-horseradish peroxidase(HPR) conjugate was added and incubated for 45 minutes.The streptavidin-HRP conjugate was captured by any biotinlabeled antibody bound to the plate. Following a wash stepto remove unbound conjugate, 3, 30, 5, 50-tetramethylbenzi-dine (TMB) substrate was added and was converted by thecaptured HRP to a colored product in proportion to theamount of HRP bound to the plate. After a short incuba-tion to allow color development, the reaction was stoppedand the intensity of the generated color was detected in amicrotiter plate reader capable of measuring 450 nm wave-length. The color development was proportional to the cap-tured test mAb or the reference mAb. All test plates werehandled identically, enabling side-by-side comparison of allsamples.
Size-exclusion chromatography
Samples were diluted to 5 mg/mL with mobile phase buffer(20 mM NaH2PO4, 150 mM NaCl, pH 7.0) prior to the analysis.SEC-HPLC was performed under non-denaturing conditionswith the Waters-2695 Alliance HPLC system on a TSKG3000SWXL column (Tosoh, Japan) with the aqueous-bufferedmobile phase. The isocratic elution profile was monitored usingUV detection at 214 nm.
SEC-multi angle light scattering
SEC-MALS was performed by HPLC on a TSKgel G3000SWXLcolumn using aqueous buffered mobile phase (20 mMNaH2PO4, 150 mM NaCl, pH 7.0). The isocratic elution profilewas monitored using MALS system, DAWN� HEREOSTMIIand Optilab rEX (Wyatt Technologies). Molecular weight andthe monomer and high molecular weight (HMW) content wasdetermined with MALS and RI detector.
Sedimentation velocity analytical ultracentrifugation
Sedimentation velocity analytical ultracentrifugation (SV-AUC) was carried out on a Beckman Coulter XLA-70 AUCinstrument at 20�C. Sample was loaded into the sample channelof AUC cells having quartz windows and 12-mm double-sectorEpon centerpieces. Matching buffer having the same composi-tion and pH as the sample were loaded into the correspondingreference channel of each cell. The centrifugation was carriedout at 20�C and 45,000 rpm. Radial scans of the concentrationprofile were collected sequentially by absorbance at 280 nm,until full sedimentation was reached. The resulting datasetswere analyzed using the program SEDFIT with a continuous c(s) distribution model, yielding best-fit distributions for thenumber of sedimenting species and the effective molecularweights. Each sample was analyzed in two separate SV-AUCspins, resulting in two analyses per sample. The resulting c(s)distribution profile was used to calculate the percentage of eachspecies and the estimated molecular weights.
Capillary sodium dodecyl sulfate gel electrophoresis
CE-SDS was performed under both non-reducing conditionsfor analysis of purity/impurities. A Beckman Coulter, PA 800capillary electrophoresis system was used with a 67 cm, 50 mmI.D. bare-fused silica capillary. Reduced CE-SDS was performedfor determination of purity as the corrected peak area % of sumof HC and LC, and non-glycosylated HCs. The samples werereduced using 2-mercaptoethanol and subjected to electropho-resis under reducing condition. Non-reduced CE-SDS was per-formed for determination of the corrected peak area % of intactIgG and non-assembled IgG molecules. The samples for non-reduced CE-SDS were alkylated using 250 mM IAM and sub-jected to electrophoresis under non-reducing condition.
Isoelectric focusing
IEF was used to determine pI values of charge variants in thesamples. Electrophoresis was performed on IsoGel agarose IEF
566 J. LEE ET AL.
plates in the pH range of pH 3 – 10 using a flatbed electropho-resis system (Multiphor II, Amersham). pI values were calcu-lated against IEF pI markers (pI range: 9.45 – 6.0) andcompared to the pI values of the reference standard. The sam-ples were focused by running the gels at 1500 V, 7 mA and25 W for 75 minutes. Following focusing, the gels were stainedwith PhastGel Blue R, dried and scanned. pI values were calcu-lated using Quantity One software (Bio-Rad).
Ion exchange chromatography
The IEC-HPLC method was used to evaluate the distribution ofcharge variants by cation exchange chromatography. TheHPLC system (Waters) was equipped with a Propac WCX 10analytical column (4 £ 250 mm) and guard column (4 £50 mm) set (Dionex) at ambient temperature. Gradient NaClelution was performed at a constant flow rate of 0.8 mL/minand UV signals were obtained at 214 nm. Peaks in the IECHPLC chromatogram were integrated, and percentage peakareas of each peak were calculated.
Oligosaccharide profile analysis by hydrophilic interactionliquid chromatography
For oligosaccharide profile analysis, N-linked glycans werereleased from the antibody using PNGase F treatment at 37�C.PNGase F-cleaved glycans were analyzed by UPLC with a FLDdetector (Waters). Released N-linked glycans by PNGase Ftreatment were extracted from deglycosylated protein solutionusing a GlycoClean H Cartridges (Prozyme). Extracted glycanswere labeled with 2-AB labeling reagent, followed by removalof excess labeling reagent using a GlycoClean S Cartridges(Prozyme). Finally, 2-AB-labeled N-linked glycans were ana-lyzed by normal phase chromatography with a BEH glycan col-umn (Waters) and fluorescence detector.
N-glycan analysis by LC-MS
For structural analysis of N-linked oligosaccharide at Asn 300by LC-MS (AB SCIEX), trypsin-digested peptides prepared forpeptide mapping were evaluated. Extracted ion chromatogramswere used to quantify each oligosaccharide species. The per-centage calculation was based on each glycosylation site. Foreach site, all detectable oligosaccharide structures were counted.
Sialic acid analysis
Sialic acids were released from the antibody by mild acidhydrolysis (0.1 M HCl) for 1 hr at 80�C. A Zorbax Extend-C18column was used at a flow rate of 1.0mL/min and followinggradient conditions using 7% methanol/ 93% water (mobilephase A) and 7% methanol/93% acetonitrile (mobile phase B);5% B initially for 35 min, increased to 100% B in 0.1 min, fol-lowed by 9.9 min isocratic hold. Initial conditions were restoredin 0.1 min and held for an additional 14.9 min to ensure col-umn equilibration. The chromatography was performed on aWaters HPLC system with fluorescent detector (FLD). Thesialic acid content was quantified based on the response of sialicacid standards (NANA) relative to an internal standard [2-
keto-3-deoxy-D-glycero-D-galacto-nononic acid; deaminatedneuraminic acid]. The results were reported as molar ratios(sialic acid/protein, mole/mole).
Glycation analysis
Samples were digested with PNGase F (1,500 U/mg treatment,at 37�C for 16 hours) for removing N-glycan, and then reducedwith DTT followed by LC-ES-MS analysis using an Agilent1200 HPLC coupled online to an Agilent 6530 Q-TOF massspectrometer. The m/z (mass/charge) data were collected from900 to 4,000 m/z at a scan rate of 1 spectrum per second fol-lowed by deconvolution of the mass spectra to intensity versusmolecular mass for both HC and LC. The percentage calcula-tion was based on deconvoluted spectra of each chain. For thedetermination of % glycation in the LC, area of glycated LCwas divided by sum of both areas from native and glycated LC.For the determination of % glycation in HC, area of glycatedHC was divided by sum of both areas from native and glycatedHC.
Spiking study with highly fragmented CT-P6 sample
Various levels (1.63 – 7.54%) of fragmented CT-P6 sampleswere generated by spiking H2L1 enriched product. The level oftotal fragmentation in each sample was confirmed by non-reduced CE-SDS. The correlation between fragmentation levelversus the relative anti-proliferation activity, FcgRIIIa-V andFcRn binding affinities were subsequently investigated.
Spiking study with non-glycosylated CT-P6 sample
Various levels (0.58 – 37.53%) of non-glycosylated CT-P6 sam-ples were generated by spiking fully aglycosylated product, pre-pared by treatment with PNGase F treatment, into controlsample (0.58% non-glycosylation). The level of non-glycosy-lated product in each sample was confirmed by reduced CE-SDS. Correlation between non-glycosylation level versus therelative anti-proliferation activity, FcgRIIIa-V and FcRn bind-ing affinities were subsequently investigated.
HER2 binding affinity
HER2 ELISA was performed for EU-Herceptin�, CT-P6 drugproduct and US-Herceptin� to measure the binding affinity toHER2. Recombinant human ErbB2/Fc chimera (R&D systems,Cat. No. 1129-ER) was coated onto the 96-well plate and over-night incubated at 2-8�C. Samples were serially diluted from500 ng/mL to 0.229 ng/mL (3-fold dilution, 8 points) andtreated to the coated plates. The bound antibody was measuredusing HRP-conjugated goat anti-human gamma chain detec-tion antibody (Sigma, Cat. No. A7164) followed by TMB treat-ment. After the stop reaction with sulfuric acid, the opticaldensity values were measured at 450 nm / 650 nm. The relativeHER2 binding affinity of sample was evaluated as (EC50 valueof CT-P6 in-house reference standard / EC50 value of the sam-ples £ 100) using a 4-PL curve model by GraphPad Prism�
software.
MABS 567
Cell-based HER2 binding affinity
HER2 CELISA was performed for EU-Herceptin�, CT-P6 drugproduct and US-Herceptin� to measure the binding affinity oftrastuzumab against HER2-overexpressing human breast can-cer cell line, SK-BR3 cells were propagated onto the 96-wellplate. After 2 days incubation, cells were fixed with 3.7% form-aldehyde and then blocked with diluent buffer. Six-fold serialdiluted sample (from 50,000 ng/mL to 0.18 ng/mL) was treatedto the fixed cells. After binding, the cells were washed and sub-sequently incubated with HRP-conjugated anti-human kappaLC detection antibody (Sigma, Cat. No. A7164). TMB wastreated to each well after the stop reaction with sulfuric acid,then the plate was measured at 450 nm / 650 nm. The relativecell-based HER2 binding affinity was determined using 4-PLlogistic curve model by PLA 3.0 software.
In vitro bioactivity (anti-proliferation activity using BT-474cell)
The in vitro bioactivity assay was performed for EU-Herceptin�, CT-P6 drug product and US-Herceptin� to mea-sure the anti-proliferation activity against the human breastcancer cell line BT474, which overexpress HER2 on the cell sur-face. These cells were incubated with EU-Herceptin�, CT-P6drug product and US-Herceptin� at concentrations from 4,000ng/mL to 7.813 ng/mL (serially 2-fold dilution, 10 points).After 5 days incubation, the bioactivity was determined bymeasuring BT-474 cell metabolic activity as an indicator of cellviability using CCK-8 assay (O.D 450 nm / 650 nm). The rela-tive potency was evaluated using 4-PL curve model by PLA 3.0software.
C1q binding affinity
An anti-C1q ELISA was used to evaluate the binding affinity ofEU-Herceptin�, CT-P6 drug product and US-Herceptin� toC1q complement. Samples were serially diluted from 60,000ng/mL to 27.43 ng/mL (3-fold dilution, 8 points) and coated ona microplate surface. After blocking with CaCl2 assay buffer,10 mg/mL of C1q (AbD Serotec, Cat. No. 2221-5504) wastreated for 2 hours, and the bound C1q was measured usinganti-C1q-HRP conjugate (AbD Serotec, Cat. No. 2221-5004P)followed by TMB treatment. After the stop reaction with sulfu-ric acid, the optical density values were measured at 450 nm /650 nm. The relative C1q binding affinity of samples was evalu-ated as (EC50 value of CT-P6 in-house reference standard /EC50 value of the samples £ 100) using a 4-PL curve model byGraphPad Prism� software.
Fcg receptor binding affinity
The Fcg receptor binding affinity assays were performed usingSPR to evaluate the binding affinity of EU-Herceptin�, CT-P6drug product and US-Herceptin� to Fc receptors. Prior to theassay, each Fc receptor was diluted in 10 mM sodium acetatebuffer, pH 5.0 and immobilized on the BIAcore CM5 chipusing an amine coupling reaction. Any unstable immobilizedFc receptor was removed by at least 3 cycles (6 cycles for
FcgRIIIa-F) of pre-run solution. Samples were serially dilutedin HBS-EP buffer, pH 7.4, (HBS-EP, pH 6.0 for FcRn) andbinding of the sample was measured in real time by the changein refractive index. The chip was regenerated in each cycleusing regeneration solution appropriate for the Fc receptor.The relative binding affinity was evaluated as (KD value of CT-P6 in-house reference standard / KD value of the samples £100) using steady state model by BIAevaluation software.
Antibody-dependent cell-mediated cytotoxicity
ADCC activity of EU-Herceptin�, CT-P6 drug product andUS-Herceptin�, which is mediated by effector cells throughFcgR binding, was assessed using the calcein-AM release assay.The human breast cancer cell line SK-BR3, which over-expresses HER2 on the cell surface, was the target cell, andhuman PBMCs were used as the effector cells. Target cell wasincubated with EU-Herceptin�, CT-P6 drug product and US-Herceptin� and PBMCs at E:T ratio (Effector cell:Target cell)of 16:1. Then, the cell cytotoxicity was measured by fluores-cence values released calcein-AM and reported as ADCC activ-ity and the relative ADCC activity of EU-Herceptin�, CT-P6drug product and US-Herceptin� was determined as mean rel-ative ADCC activity of each cytotoxicity at 3 concentrations(0.5, 1.3 and 3.2 ng/mL) within the linear range of ADCCresponse compared with that of CT-P6 in-house referencestandard.
Abbreviation
ADCC antibody-dependent cell-mediated cytotoxicity2-AB 2-aminobenzamideCD circular dichroismCDC complement-dependent cytotoxicityCELISA cell-based enzyme linked immunosorbent assayCE-SDS capillary sodium dodecyl sulfate gel
electrophoresisCHO Chinese hamster ovaryCI confidence intervalDSC differential scanning calorimetryDTT dithiothreitolELISA enzyme linked immunosorbent assayEM equivalence marginFT-IR Fourier transform infra-redHER2 human epidermal growth factor receptor 2HILIC hydrophilic interaction liquid chromatographyHPLC high performance liquid chromatographyHMW high molecular weightHRP horseradish peroxidaseIAM iodoacetamideIEC ion exchange chromatographyIEF isoelectric focusingLC-MS-ESI liquid chromatography mass spectrometry elec
trospray ionizationLMW low molecular weightMALS multi-angle light scatteringNANA N-acetylneuraminic acidQR quality rangeRP reference product
568 J. LEE ET AL.
SEC size-exclusion chromatographySPR surface plasmon resonanceSV-AUV sedimentation velocity analytical
ultracentrifugationTMB 3, 30, 5, 50-tetramethylbenzidineUPLC ultra-performance liquid chromatography
Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.
Funding
This work was suppported by the Celltrion Inc.
ORCID
Jihun Lee http://orcid.org/0000-0002-8803-7689Jin Soo Bae http://orcid.org/0000-0001-7484-9048
References
1. European Medicines Agency, Committee for Medicinal Products forHuman Use. Guideline on similar biological medicinal products[Internet]. London, UK: European Medicines Agency; [Adopted on2014 Oct 23; cited 2017 Dec 15]. Available from: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2014/10/WC500176768.pdf.
2. European Medicines Agency, Committee for Medicinal Products forHuman Use. Guideline on similar biological medicinal products con-taining biotechnology-derived proteins as active substance: Qualityissues (revision 1) [Internet]. London, UK: European MedicinesAgency; [Adopted on 2014 May 22; cited 2017 Dec 13]. Availablefrom: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2014/06/WC500167838.pdf.
3. US Food and Drug Administration. Guidance for Industry: ScientificConsideration in Demonstrating Biosimilarity to a Reference Product;2015 April [accessed 2017 Sept 15]. https://www.fda.gov/downloads/drugs/guidances/ucm291128.pdf.
4. US Food and Drug Administration. Guidance for Industry: QualityConsiderations in Demonstrating Biosimilarity of a Therapeutic ProteinProduct to a Reference Product; 2015 April [accessed 2017 Sept 15].https://www.fda.gov/downloads/drugs/guidances/ucm291128.pdf.
5. Liu J, Eris T, Li C, Cao S, Kuhns S. Assessing Analytical Similarity ofProposed Amgen Biosimilar ABP 501 to Adalimumab. BioDrugs.2016;30:321–38. doi:10.1007/s40259-016-0184-3. PMID:27461107.
6. Hong J, Lee Y, Lee C, Eo S, Kim S, Lee N, Park J, Park S, Seo D, JeongM, et al. Physicochemical and biological characterization of SB2, abiosimilar of Remicade(R) (infliximab). MAbs. 2017;9:364–82.doi:10.1080/19420862.2016.1264550. PMID:28005456.
7. Lee C, Jeong M, Lee JJ, Seo S, Cho SC, Zhang W, Jaquez O. Glycosyla-tion profile and biological activity of Remicade(R) compared withFlixabi(R) and Remsima(R). MAbs. 2017;9:968–77. doi:10.1080/19420862.2017.1337620. PMID:28640663.
8. Magnenat L, Palmese A, Fremaux C, D’Amici F, Terlizzese M, RossiM, Chevalet L. Demonstration of physicochemical and functionalsimilarity between the proposed biosimilar adalimumab MSB11022and Humira(R). MAbs. 2017;9:127–39. doi:10.1080/19420862.2016.1259046. PMID:27854156.
9. Hassett B, Scheinberg M, Castaneda-Hernandez G, Li M, Rao URK,Singh E, Mahgoub E, Coindreau J, O’Brien J, Vicik SM, et al. Variabil-ity of intended copies for etanercept (Enbrel(R)): Data on multiplebatches of seven products. MAbs. 2018;10:166–76. doi:10.1080/19420862.2017.1387346. PMID:29020508.
10. Hassett B, Singh E, Mahgoub E, O’Brien J, Vicik SM, Fitzpatrick B.Manufacturing history of etanercept (Enbrel((R))): Consistency of
product quality through major process revisions. MAbs. 2018;10:159–65. doi:10.1080/19420862.2017.1388483. PMID:29020515.
11. European Medicines Agency. Herzuma: Summary of Opinion (Initialauthorisation). Committee for Medicinal Products for Human Use(CHMP). 2017.
12. US Food and Drug Administration. FDA approved Drug Products;[accessed 2017 Dec 5]. Available from: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?eventDoverview.process&applnoD103792.
13. European Medicines Agency, Committee for Medicinal Products forHuman Use. Herceptin: European public assessment report (EPAR)summary for the public [Internet]. London, UK: European MedicinesAgency; [Updated in 2013 July; cited 2017 Dec 3]. Available from:http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Summary_for_the_public/human/000278/WC500049819.pdf.
14. Arteaga CL. Trastuzumab, an appropriate first-line single-agent ther-apy for HER2-overexpressing metastatic breast cancer. Breast CancerRes. 2003;5:96–100. doi:10.1186/bcr574. PMID:12631388.
15. Clynes RA, Towers TL, Presta LG, Ravetch JV. Inhibitory Fc receptorsmodulate in vivo cytotoxicity against tumor targets. Nat Med.2000;6:443–6. doi:10.1038/74704. PMID:10742152.
16. Collins DM, O’Donovan N, McGowan PM, O’Sullivan F, Duffy MJ,Crown J. Trastuzumab induces antibody-dependent cell-mediatedcytotoxicity (ADCC) in HER-2-non-amplified breast cancer cell lines.Ann Oncol. 2012;23:1788–95. doi:10.1093/annonc/mdr484.PMID:22056974.
17. Junttila TT, Parsons K, Olsson C, Lu Y, Xin Y, Theriault J, Crocker L,Pabonan O, Baginski T, Meng G, et al. Superior in vivo efficacy of afu-cosylated trastuzumab in the treatment of HER2-amplified breast can-cer. Cancer Res. 2010;70:4481–9. doi:10.1158/0008-5472.CAN-09-3704. PMID:20484044.
18. Petricevic B, Laengle J, Singer J, Sachet M, Fazekas J, Steger G, BartschR, Jensen-Jarolim E, Bergmann M. Trastuzumab mediates antibody-dependent cell-mediated cytotoxicity and phagocytosis to the sameextent in both adjuvant and metastatic HER2/neu breast cancerpatients. J Transl Med. 2013;11:307. doi:10.1186/1479-5876-11-307.PMID:24330813.
19. European Medicines Agency. Ontruzant: Summary of Opinion (Initialauthorisation). Committee for Medicinal Products for Human Use(CHMP). 2017.
20. US Food and Drug Administration. FDA approves Ogivri as a biosi-milar to Herceptin; [accessed 2017 Dec 15]. Available from: https://www.fda.gov/drugs/informationondrugs/approveddrugs/ucm587404.htm.
21. Damen CW, Chen W, Chakraborty AB, van Oosterhout M, MazzeoJR, Gebler JC, Schellens JH, Rosing H, Beijnen JH. Electrospray ioni-zation quadrupole ion-mobility time-of-flight mass spectrometry as atool to distinguish the lot-to-lot heterogeneity in N-glycosylation pro-file of the therapeutic monoclonal antibody trastuzumab. J Am SocMass Spectrom. 2009;20:2021–33. doi:10.1016/j.jasms.2009.07.017.PMID:19744865.
22. Beck A, Sanglier-Cianferani S, VanDorsselaer A. Biosimilar, biobetter, andnext generation antibody characterization by mass spectrometry. AnalChem. 2012;84:4637–46. doi:10.1021/ac3002885. PMID:22510259.
23. Kim S, Song J, Park S, Ham S, Paek K, Kang M, Chae Y, Seo H, KimHC, Flores M. Drifts in ADCC-related quality attributes of Herceptin(R): Impact on development of a trastuzumab biosimilar. MAbs.2017;9:704–14. doi:10.1080/19420862.2017.1305530. PMID:28296619.
24. Schiestl M, Stangler T, Torella C, Cepeljnik T, Toll H, Grau R. Acceptablechanges in quality attributes of glycosylated biopharmaceuticals. Nat Bio-technol. 2011;29:310–2. doi:10.1038/nbt.1839. PMID:21478841.
25. US Food and Drug Administration. Guidance for Industry: StatisticalApproaches to Evaluate Analytical Similarity; 2017 Sept [accessed2017 Dec 3]. https://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm576786.pdf.
26. European Medicines Agency, Committee for Medicinal Products forHuman Use. Guideline on similar biological medicinal products contain-ing biotechnology - non-clinical and clinical issues [Internet]. London,UK: European Medicines Agency; [Adopted on 2012 May 30; cited 2017Dec 3]. Available from: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/06/WC500128686.pdf.
MABS 569
27. Molina MA, Codony-Servat J, Albanell J, Rojo F, Arribas J, Baselga J.Trastuzumab (herceptin), a humanized anti-Her2 receptor monoclo-nal antibody, inhibits basal and activated Her2 ectodomain cleavagein breast cancer cells. Cancer Res. 2001;61:4744–9. PMID:11406546.
28. Gemmete JJ, Mukherji SK. Trastuzumab (herceptin). AJNR Am J Neuro-radiol. 2011;32:1373–4. doi:10.3174/ajnr.A2619. PMID:21816914.
29. Junttila TT, Akita RW, Parsons K, Fields C, Lewis Phillips GD,Friedman LS, Sampath D, Sliwkowski MX. Ligand-independentHER2/HER3/PI3K complex is disrupted by trastuzumab and is effec-tively inhibited by the PI3K inhibitor GDC-0941. Cancer Cell.2009;15:429–40. doi:10.1016/j.ccr.2009.03.020. PMID:19411071.
30. Valabrega G, Montemurro F, Aglietta M. Trastuzumab: mechanism ofaction, resistance and future perspectives in HER2-overexpressingbreast cancer. Ann Oncol. 2007;18:977–84. doi:10.1093/annonc/mdl475. PMID:17229773.
31. Nahta R. Molecular Mechanisms of Trastuzumab-Based Treatment inHER2-Overexpressing Breast Cancer. ISRN Oncol. 2012;2012:428062.PMID:23227361.
32. She QB, Chandarlapaty S, Ye Q, Lobo J, Haskell KM, Leander KR,DeFeo-Jones D, Huber HE, Rosen N. Breast tumor cells with PI3Kmutation or HER2 amplification are selectively addicted to Akt signal-ing. PLoS One. 2008;3:e3065. doi:10.1371/journal.pone.0003065.PMID:18725974.
33. Vu T, Claret FX. Trastuzumab: updated mechanisms of action andresistance in breast cancer. Front Oncol. 2012;2:62. doi:10.3389/fonc.2012.00062. PMID:22720269.
34. Yeap WH, Wong KL, Shimasaki N, Teo EC, Quek JK, Yong HX,Diong CP, Bertoletti A, Linn YC, Wong SC. CD16 is indispensable forantibody-dependent cellular cytotoxicity by human monocytes. SciRep. 2016;6:34310. doi:10.1038/srep34310. PMID:27670158.
35. Seidel UJ, Schlegel P, Lang P. Natural killer cell mediated antibody-dependent cellular cytotoxicity in tumor immunotherapy with thera-peutic antibodies. Front Immunol. 2013;4:76. doi:10.3389/fimmu.2013.00076. PMID:23543707.
36. Nahta R, Yu D, Hung MC, Hortobagyi GN, Esteva FJ. Mechanisms ofdisease: understanding resistance to HER2-targeted therapy in humanbreast cancer. Nat Clin Pract Oncol. 2006;3:269–80. doi:10.1038/ncponc0509. PMID:16683005.
37. Gennari R, Menard S, Fagnoni F, Ponchio L, Scelsi M, Tagliabue E, Casti-glioni F, Villani L, Magalotti C, Gibelli N, et al. Pilot study of the mecha-nism of action of preoperative trastuzumab in patients with primaryoperable breast tumors overexpressing HER2. Clin Cancer Res.2004;10:5650–5. doi:10.1158/1078-0432.CCR-04-0225. PMID:15355889.
38. Carter P, Presta L, Gorman CM, Ridgway JB, Henner D, Wong WL,Rowland AM, Kotts C, Carver ME, Shepard HM. Humanization of ananti-p185HER2 antibody for human cancer therapy. Proc Natl Acad SciU S A. 1992;89:4285–9. doi:10.1073/pnas.89.10.4285. PMID:1350088.
39. Cooley S, Burns LJ, Repka T, Miller JS. Natural killer cell cytotoxicityof breast cancer targets is enhanced by two distinct mechanisms ofantibody-dependent cellular cytotoxicity against LFA-3 and HER2/neu. Exp Hematol. 1999;27:1533–41. doi:10.1016/S0301-472X(99)00089-2. PMID:10517495.
40. Lewis GD, Figari I, Fendly B, Wong WL, Carter P, Gorman C, She-pard HM. Differential responses of human tumor cell lines to anti-p185HER2 monoclonal antibodies. Cancer Immunol Immunother.1993;37:255–63. doi:10.1007/BF01518520. PMID:8102322.
41. Stockmeyer B, Beyer T,NeuhuberW, Repp R, Kalden JR, Valerius T, Herr-mann M. Polymorphonuclear granulocytes induce antibody-dependentapoptosis in human breast cancer cells. J Immunol. 2003;171:5124–9.doi:10.4049/jimmunol.171.10.5124. PMID:14607911.
42. Nimmerjahn F, Ravetch JV. Fcgamma receptors as regulators ofimmune responses. Nat Rev Immunol. 2008;8:34–47. doi:10.1038/nri2206. PMID:18064051.
43. Wang W, Erbe AK, Hank JA, Morris ZS, Sondel PM. NK Cell-Medi-ated Antibody-Dependent Cellular Cytotoxicity in Cancer Immuno-therapy. Front Immunol. 2015;6:368. doi:10.3389/fimmu.2015.00368.PMID:26284063.
44. Nimmerjahn F, Gordan S, Lux A. FcgammaR dependent mechanisms ofcytotoxic, agonistic, and neutralizing antibody activities. Trends Immunol.2015;36:325–36. doi:10.1016/j.it.2015.04.005. PMID:25981969.
45. Herbrand U. Antibody-Dependent Cellular Phagocytosis: The Mecha-nism of Action That Gets No Respect A Discussion About ImprovingBioassay Reproducibility. BioProcess J. 2016;15:26–9. doi:10.12665/J151.Herbrand.
46. Tada M, Ishii-Watabe A, Suzuki T, Kawasaki N. Development of acell-based assay measuring the activation of FcgammaRIIa for thecharacterization of therapeutic monoclonal antibodies. PLoS One.2014; 9:e95787. doi:10.1371/journal.pone.0095787. PMID:24752341.
47. Suzuki T, Ishii-Watabe A, Tada M, Kobayashi T, Kanayasu-Toyoda T,Kawanishi T, Yamaguchi T. Importance of neonatal FcR in regulatingthe serum half-life of therapeutic proteins containing the Fc domainof human IgG1: a comparative study of the affinity of monoclonalantibodies and Fc-fusion proteins to human neonatal FcR. J Immunol.2010;184:1968–76. doi:10.4049/jimmunol.0903296. PMID:20083659.
48. Cho HS, Mason K, Ramyar KX, Stanley AM, Gabelli SB, Denney DW,Jr., Leahy DJ. Structure of the extracellular region of HER2 alone andin complex with the Herceptin Fab. Nature. 2003;421:756–60.doi:10.1038/nature01392. PMID:12610629.
49. Harris RJ, Kabakoff B, Macchi FD, Shen FJ, Kwong M, Andya JD,Shire SJ, Bjork N, Totpal K, Chen AB. Identification of multiple sour-ces of charge heterogeneity in a recombinant antibody. J ChromatogrB Biomed Sci Appl. 2001;752:233–45. doi:10.1016/S0378-4347(00)00548-X. PMID:11270864.
50. Worobec A, Rosenberg AS. A Risk-Based Approach to Immunogenic-ity Concerns of Therapeutic Protein Products, Part 1: ConsideringConsequences of the Immune Response to a Protein. BioPharm Inter-national. 2004;17:22–6.
51. Worobec A, Rosenberg AS. A Risk-Based Approach to Immunogenic-ity Concerns of Therapeutic Protein Products, Part 2: Consideringhost-specific and product-specific factors impacting immunogenicity.BioPharm International. 2004;17:34–42.
52. Rosenberg AS. Effects of protein aggregates: an immunologic per-spective. AAPS J. 2006;8:E501–7. doi:10.1208/aapsj080359. PMID:17025268.
53. Hermeling S, Aranha L, Damen JM, Slijper M, Schellekens H, Crom-melin DJ, Jiskoot W. Structural characterization and immunogenicityin wild-type and immune tolerant mice of degraded recombinanthuman interferon alpha2b. Pharm Res. 2005;22:1997–2006.doi:10.1007/s11095-005-8177-9. PMID:16184451.
54. Sauerborn M, Brinks V, Jiskoot W, Schellekens H. Immunologicalmechanism underlying the immune response to recombinant humanprotein therapeutics. Trends Pharmacol Sci. 2010;31:53–9.doi:10.1016/j.tips.2009.11.001. PMID:19963283.
55. Cordoba AJ, Shyong BJ, Breen D, Harris RJ. Non-enzymatic hingeregion fragmentation of antibodies in solution. J Chromatogr B AnalytTechnol Biomed Life Sci. 2005;818:115–21. doi:10.1016/j.jchromb.2004.12.033. PMID:15734150.
56. Khawli LA, Goswami S, Hutchinson R, Kwong ZW, Yang J, Wang X,Yao Z, Sreedhara A, Cano T, Tesar D, et al. Charge variants in IgG1:Isolation, characterization, in vitro binding properties and pharmaco-kinetics in rats. MAbs. 2010;2:613–24. doi:10.4161/mabs.2.6.13333.PMID:20818176.
57. Ravetch JV, Bolland S. IgG Fc receptors. Annu Rev Immunol.2001;19:275–90. doi:10.1146/annurev.immunol.19.1.275. PMID:11244038.
58. Jefferis R. Recombinant antibody therapeutics: the impact of glycosyl-ation on mechanisms of action. Trends Pharmacol Sci. 2009;30:356–62. doi:10.1016/j.tips.2009.04.007. PMID:19552968.
59. Zheng K, Bantog C, Bayer R. The impact of glycosylation on monoclo-nal antibody conformation and stability. MAbs. 2011;3:568–76.doi:10.4161/mabs.3.6.17922. PMID:22123061.
60. Shields RL, Lai J, Keck R, O’Connell LY, Hong K, Meng YG, WeikertSH, Presta LG. Lack of fucose on human IgG1 N-linked oligosaccha-ride improves binding to human Fcgamma RIII and antibody-depen-dent cellular toxicity. J Biol Chem. 2002;277:26733–40. doi:10.1074/jbc.M202069200. PMID:11986321.
61. Shinkawa T, Nakamura K, Yamane N, Shoji-Hosaka E, Kanda Y,Sakurada M, Uchida K, Anazawa H, Satoh M, Yamasaki M, et al. Theabsence of fucose but not the presence of galactose or bisecting N-ace-tylglucosamine of human IgG1 complex-type oligosaccharides showsthe critical role of enhancing antibody-dependent cellular cytotoxicity.
570 J. LEE ET AL.
J Biol Chem. 2003;278:3466–73. doi:10.1074/jbc.M210665200.PMID:12427744.
62. Suzuki E, Niwa R, Saji S, Muta M, Hirose M, Iida S, Shiotsu Y, SatohM, Shitara K, Kondo M,. A nonfucosylated anti-HER2 antibody aug-ments antibody-dependent cellular cytotoxicity in breast cancerpatients. Clin Cancer Res. 2007;13:1875–82. doi:10.1158/1078-0432.CCR-06-1335. PMID:17363544.
63. Mori K, Kuni-Kamochi R, Yamane-Ohnuki N, Wakitani M, YamanoK, Imai H, Kanda Y, Niwa R, Iida S, Uchida K, et al. Engineering Chi-nese hamster ovary cells to maximize effector function of producedantibodies using FUT8 siRNA. Biotechnol Bioeng. 2004;88:901–8.doi:10.1002/bit.20326. PMID:15515168.
64. Okazaki A, Shoji-Hosaka E, Nakamura K, Wakitani M, Uchida K,Kakita S, Tsumoto K, Kumagai I, Shitara K. Fucose depletion fromhuman IgG1 oligosaccharide enhances binding enthalpy and associa-tion rate between IgG1 and FcgammaRIIIa. J Mol Biol.2004;336:1239–49. doi:10.1016/j.jmb.2004.01.007. PMID:15037082.
65. Jenkins N, Parekh RB, James DC. Getting the glycosylation right:implications for the biotechnology industry. Nat Biotechnol.1996;14:975–81. doi:10.1038/nbt0896-975. PMID:9631034.
66. Zhang L, Luo S, Zhang B. Glycan analysis of therapeutic glycopro-teins. MAbs. 2016;8:205–15. doi:10.1080/19420862.2015.1117719.PMID:26599345.
67. Reusch D, Tejada ML. Fc glycans of therapeutic antibodies as criticalquality attributes. Glycobiology. 2015;25:1325–34. doi:10.1093/glycob/cwv065. PMID:26263923.
68. Goetze AM, Liu YD, Zhang Z, Shah B, Lee E, Bondarenko PV, FlynnGC. High-mannose glycans on the Fc region of therapeutic IgG anti-bodies increase serum clearance in humans. Glycobiology.2011;21:949–59. doi:10.1093/glycob/cwr027. PMID:21421994.
69. Alessandri L, Ouellette D, Acquah A, Rieser M, Leblond D, SaltarelliM, Radziejewski C, Fujimori T, Correia I. Increased serum clearanceof oligomannose species present on a human IgG1 molecule. MAbs.2012;4:509–20. doi:10.4161/mabs.20450. PMID:22669558.
70. Johnson KA, Paisley-Flango K, Tangarone BS, Porter TJ, Rouse JC.Cation exchange-HPLC and mass spectrometry reveal C-terminalamidation of an IgG1 heavy chain. Anal Biochem. 2007;360:75–83.doi:10.1016/j.ab.2006.10.012. PMID:17113563.
71. Visser J, Feuerstein I, Stangler T, Schmiederer T, Fritsch C, Schiestl M.Physicochemical and functional comparability between the proposed
biosimilar rituximab GP2013 and originator rituximab. BioDrugs.2013;27:495–507. doi:10.1007/s40259-013-0036-3. PMID:23649935.
72. Hodoniczky J, Zheng YZ, James DC. Control of recombinantmonoclonal antibody effector functions by Fc N-glycan remodel-ing in vitro. Biotechnol Prog. 2005;21:1644–52. doi:10.1021/bp050228w. PMID:16321047.
73. Boyd PN, Lines AC, Patel AK. The effect of the removal of sialic acid,galactose and total carbohydrate on the functional activity of Cam-path-1H. Mol Immunol. 1995;32:1311–8. doi:10.1016/0161-5890(95)00118-2. PMID:8643100.
74. Prang N, Preithner S, Brischwein K, Goster P, Woppel A, Muller J,Steiger C, Peters M, Baeuerle PA, da Silva AJ. Cellular and comple-ment-dependent cytotoxicity of Ep-CAM-specific monoclonal anti-body MT201 against breast cancer cell lines. Br J Cancer.2005;92:342–9. doi:10.1038/sj.bjc.6602310. PMID:15655555.
75. Jefferis R. Glycosylation of recombinant antibody therapeutics. Bio-technol Prog. 2005;21:11–6. doi:10.1021/bp040016j. PMID:15903235.
76. Quan C, Alcala E, Petkovska I, Matthews D, Canova-Davis E, TaticekR, Ma S. A study in glycation of a therapeutic recombinant humanizedmonoclonal antibody: where it is, how it got there, and how it affectscharge-based behavior. Anal Biochem. 2008;373:179–91. doi:10.1016/j.ab.2007.09.027. PMID:18158144.
77. Esteva FJ, Lee SJ, Stebbing J, Winkle PJ, Wood-Horrall RN, Lee SY.Phase I clinical trial comparing PK and safety of trastuzumab and itsbiosimilar candidate CT-P6 [Abstract]. The Breast. 2017;32S1:S68–S9.doi:10.1016/S0960-9776(17)30232-1.
78. Stebbing J, Baranau Y, Baryash V, Manikhas A, Moiseyenko V, Dzag-nidze G, Zhavrid E, Boliukh D, Stroyakovskii D, Pikiel J, et al. CT-P6compared with reference trastuzumab for HER2-positive breast can-cer: A randomised, double-blind, active-controlled, phase 3 equiva-lence trial. Lancet Oncol. 2017;18:917–28. doi:10.1016/S1470-2045(17)30434-5. PMID:28592386.
79. Tsong Y, Dong X, Shen M. Development of Statistical Methods forAnalytical Similarity Assessment. J Biopharm Stat. 2015. doi:10.1080/10543406.2015.1092038.
80. Tsong Y, Dong X, Shen M. Development of statistical methods foranalytical similarity assessment. J Biopharm Stat. 2017;27:197–205.doi:10.1080/10543406.2016.1272606. PMID:27977326.
81. PASS13. Power Analysis and Sample Size Software. Kaysville, Utah,USA: NCSS, 2014.
MABS 571